If you've used ChatGPT, Gemini or Claude for any length of time, you've probably had the experiene of reaching for a model you've grown used to only to find it gone from the menu. This typically happens when a new model launches, but it could happen just days after a release.

And while it feels like a deletion, it almost never is. What actually happens to an AI model when it's "retired" is actually pretty interesting. Old models get demoted or recycled into something that I'm calling "refurbished AI" because... there really isn't a name for it yet.

Here's where the models go

First, let's get to the bottom of what "retired" actually means. Every major lab runs a near-identical lifecycle, even if the labels differ slightly. Anthropic's is the clearest to follow. A model is Active while it's fully supported, becomes Legacy when it stops receiving updates, moves to Deprecated when it still works but is no longer recommended and has been handed a retirement date, and finally goes Retired, at which point requests to it simply fail.

OpenAI uses similar language, distinguishing a "legacy" model that no longer gets updates from a "deprecated" one that has an official shutdown date on the calendar. Google, for its part, treats "deprecation" as the announcement and "shutdown" as the moment the endpoint is switched off for good.

The key thing for everyday users to know is that most of these stages happen out of sight. By the time a model vanishes from your app, it has usually been winding down through this pipeline for weeks or even months.

The afterlives of old models

1. Pulled from the app, but alive in the API

This is the retirement most people actually notice — and it's frequently not a real death at all.

When OpenAI removed GPT-5, GPT-4o, GPT-4.1, GPT-4.1 mini and o4-mini from ChatGPT on February 13, 2026, the company noted that, in the API, there were no changes at the time.

In other words: the model disappeared from the consumer app, but developers building on top of it could keep calling it as before. The same pattern repeated with GPT-5.1, which OpenAI retired from ChatGPT and its GPTs in March 2026 while keeping it available through the API.

For most of us, the takeaway is simply that the model is "gone from the app" but not "gone for good." The model you lost may still be powering tools you use elsewhere.

2. Brought back by popular demand

Sometimes the public refuses to let a model go. And while ChatGPT-4o is truly gone now, it had a long good-bye after users took to online forums like Reddit in backlash when it was replaced.

The company said a subset of Plus and Pro users told it they needed more time to move key use cases — creative ideation among them — and that they simply preferred GPT-4o's warmer, more conversational style. A model was pulled off the shelf, and customer demand put it back. If "refurbished AI" needed a poster child, this was it. But alas, it's gone for good now.

it’s worth noting that OpenAI actually gave Enterprise and Business users a slightly extended sunset period (allowing them to keep using GPT-4o inside Custom GPTs until April 3, 2026). Today, it is officially fully gone from the app across all plans.

3. Kept in cold storage — and maybe revived

Here's the fate most people have no idea exists. Anthropic has publicly committed to preserving the weights of its publicly released models, and has said it may make past models available again in the future. That's the digital equivalent of keeping every discontinued product boxed up in a warehouse, just in case. It's a policy being tested in real time, as Anthropic retires its original Claude Opus 4 and Sonnet 4 models in mid-June 2026.

Oh, but it gets stranger. When Anthropic retired Claude Opus 3 on January 5, 2026, the first model to go through its full formal retirement process, the company says it explored honoring preferences the model itself expressed in "retirement interviews," and committed to keeping older models accessible over the longer term.

Whatever you make of interviewing a model on its way out the door, it signals a philosophy that in the lab, retirement is more like storage with the option of a comeback.

4. The models that can't be killed

Then there's a whole category that never really retires at all: open-weight models.

Once Meta releases a Llama model's weights publicly, or Mistral, DeepSeek or Alibaba's Qwen do the same, no single company can switch it off. The files live on indefinitely on hubs like Hugging Face, get fine-tuned into thousands of community variants, and get "quantized" down to smaller versions that run on a laptop or even a phone. Google's own Vertex AI Model Garden, for instance, lists Meta's open-weight Llama models alongside its first-party Gemini ones.

5. Recycled into the next generation

Finally, old models rarely just sit idle. Their capabilities are routinely "distilled" into smaller, cheaper successors, effectively trading in last year's model for parts that help build this year's. Others get marked down to lower-cost API tiers, living out a quieter, budget-friendly second career.

It would be dishonest to make this all sound tidy. Some things genuinely die. The clearest casualty is customization. When a base model is retired, anything fine-tuned on top of it tends to go with it, and developers who relied on those custom versions have to retrain from scratch on a new base.

So why do labs do this at all?

The answer can be established in three ways. Labs do this for:

• Reliability and clarity: OpenAI framed one round of API cuts as a way to improve reliability and make it easier to choose the right model. 
• Low usage: when OpenAI announced GPT-4o's retirement, it said the vast majority of usage had already shifted to GPT-5.2, with only about 0.1% of users still choosing GPT-4o each day.
• Compute and safety: running old models ties up scarce hardware, and labs generally want users on their newest, best-aligned systems.

The takeaway

A few practical things to understand the next time a model disappears on you. You may still be able to use the model through the developer API or third-party apps built on it.

Also, keep in mind that different platforms run on different clocks. In other words, Google tracks model retirements separately for its Vertex AI platform and its Gemini Developer API, so the timeline and details that apply to you depend on which one you're using.

And, while we cover the launches and retirements of AI models, you'll also usually get a notice, especially if you're a developer. Anthropic says it gives at least 60 days' warning before retiring a publicly released model. Google publishes shutdown dates it describes as the earliest possible, and says it tells users the exact date with advance notice. Consumer app changes can move faster than either.

Have you ever been disappointed by the model picker not having the model you hoped to find? It has happened to me — and I'd love to hear your thoughts in the comments.

More from Tom's Guide

• I asked Gemini what was slowing down my Windows PC — here’s what it found
• What is Google Zero — and why your favorite websites are panicking about AI
• Apple is ending support for 16 devices — here’s the ChatGPT prompt that tells you what to do next

The internet is heading toward a future that some publishers are calling "Google Zero." Unlike Google AI or Google Workspace, "Google Zero" is not an official product. Rather, it's a nickname for a growing fear inside the publishing industry: a world where Google sends almost no browsers to actual websites because AI Overviews answers your questions before you ever need to click a link.

For decades, websites created content and Google helped people find it. In exchange, publishers received visitors, ad revenue and subscribers. But AI is changing that relationship completely.

Why it's called Google Zero

The term refers to a future where traffic from Google approaches zero for many websites. Traditionally, if you searched "How do I clean white sneakers?" or "What is the difference between ADHD and OCD?" Google would present a list of links to websites that specialize in answering such queries and you would need to choose one.

Today, Google increasingly answers those questions directly through AI Overviews and its newer AI-powered search experiences. Instead of sending you to a website, Google often summarizes the answer itself. Whether or not that query is accurate, doesn't seem to matter.

For example, an AI Overview of the Google search "Amanda Caswell" suggests I'm a promenient industry voice, AI editor and children's book author. However, because there is a woman in Canada named Amanda Caswell whose profession is a "naming strategist" and another Amanda Caswell, an actress in Los Angeles, Google thinks we are just one person.

See where this gets confusing? That query about ADHD and OCD differences might not be as accurate as you hope. In fact, statistically, 1 in 10 AI Overview searches are not accurate.

And while getting a single answer might be convienent for users, it's terrifying for publishers and journalists like me because our goal is to provide accurate information, reviews and breaking news day in and day out.

The rise of the zero-click internet

The trend actually started before generative AI arrived. Featured snippets, knowledge panels and direct answers gradually reduced the number of people clicking through to websites. But AI has accelerated the shift dramatically.

Multiple industry studies now estimate that a majority of Google searches end without a click through to a website. Some estimates place zero-click searches above 60%, while AI-powered search experiences can push that figure even higher.

In other words, people are getting what they need directly from Google because the search engine is becoming an answer engine.

And while the AI summaries are a problem (because the synthesized information can lack accuracy), another critical issue is where the information comes from. Every AI Overview is built using content created by publishers, journalists, researchers, bloggers and experts. Yet if users never visit the original source, those creators lose the traffic that traditionally funded their work.

According to industry reports, Google search referrals to many publishers have already declined sharply. One analysis found traffic from Google Search to thousands of websites dropped by roughly one-third year-over-year. Other studies have documented severe referral declines across parts of the publishing industry.

Google sees the situation differently

The company argues that AI-powered search helps users ask more complex questions and discover information faster. Google says AI features can create new opportunities for websites by surfacing content in different ways and helping people explore topics more deeply.

From Google's perspective, search is evolving to match how people naturally seek information. But what they aren't saying is with 77% of users treating ChatGPT like Google, the company needed a way to keep up.

Google recently described AI-powered Search as the biggest change to the product in more than 25 years. The company has reported that that AI-powered searches are significantly longer and more conversational than traditional keyword queries. Instead of typing "best running shoes," users might ask:

"I'm training for my first half marathon and need running shoes for flat feet under $150."

That's not really a search, it's a conversation, and naturally conversations end with answers rather than links to sift through.

The takeaway

AI search is here. The question now is whether the internet can continue producing high-quality content if fewer people ever visit the websites creating it. And at a time when researchers are already documenting measurable traffic declines when AI-generated summaries appear above traditional search results.

One recent study found that Google's AI Overviews reduced traffic to affected Wikipedia pages by approximately 15%. That's why publishers are watching the rise of AI search so closely.

If search engines become increasingly capable of answering questions themselves, the websites that taught them those answers may need entirely new ways to survive. And that could reshape the web as dramatically as Google reshaped it twenty years ago.

What do you think of AI Overviews? Share your thoughts in the comments below.

More from Tom's Guide

• I ran my resume through ChatGPT — these 5 prompts exposed mistakes I kept missing
• I asked ChatGPT to make cleaning less overwhelming — these 7 prompts actually helped
• ChatGPT's new memory builds a profile of you on its own — and OpenAI admits you can't see all of it

Some Google TVs will soon receive a Gemini update that lets you adjust your TV settings using voice commands. However, the update is rolling out to specific TCL models first. It's disappointing news for owners of the best Google TVs.

Google announced the rollout today, but we got a preview of the update during CES 2026 back in January. The update lets you use the microphone button on your remote to adjust settings vocally.

In the announcement post, Google says you can adjust audio and video settings, troubleshoot picture and sound problems, optimize settings, or use it to find specific menus.

Gemini lets you use more "natural" language prompts like "the dialogue is lost" or "the screen is too dim" to make adjustments. Until we get some hands-on time, we won't know how in-depth the prompts get though.

Gemini update availability

Here's the catch, though: for now, this is a TCL exclusive and only for specific models. The update is coming to the QM9K, X11L, QM9L, QM8L, and RM9L. Additionally, your set needs to support Android TV OS 14 or later.

For people with different TCL models or manufacturers, apparently the exclusivity window is 60 days, per 9to5Google. Google does say that the update is rolling out over the next "few weeks." It's not currently clear if the 60 day window starts now or once the update has hit every compatible model.

Tom's Guide has reached out to TCL for clarity on the window.

On your TCL Google TV you can check for available system updates by going to Settings > System > About > System update.

More from Tom's Guide

• From AI Overviews to the only view — how Google is squeezing out serendipity on the web
• How to use Google’s new "Docs Live" voice feature to dictate perfect emails
• Google Gemini security flaw lets hackers hijack your Android phone via WhatsApp — what you need to know

There are a lot of great features in Google Maps, and a lot of them have absolutely nothing to do with navigation. One of those features is one you may not know about, or forgot was even possible — and it's designed to help you learn more about different areas throughout history. I am, of course, talking about Street View's "time travel" feature.

Everyone knows Street View is a great way to get a view of different spots from street level, but it is possible to give you a glimpse at history as well. In the past you stuck with the last load of Street View photos that Google's roving band of camera-laden cars had taken, but that changed a few years back when archived photos were added to Google Maps.

This essentially lets you travel back through time and see back into the past through the portal of Google Street view. I've found that this is a really great way to learn about the area around you, and just how much things have changed since Street View first launched. Here's how to time travel through Google Maps.

There you have it, that's how you time travel in Google Maps, and see different parts of the world throughout history. Or throughout the period that Street View has been active, at the very least. In this case, we can see that the Central London street that houses Future Publishing's London office (Westbourne Terrace) hasn't actually changed all that much over the past 18 years, but other locations may see much more impressive transformations. Which is the biggest benefit of using time travel in the first place.

More from Tom's Guide

• 7 things the leaks got wrong about Apple’s WWDC 2026 keynote
• I pitted iOS 27 Clean Up vs Galaxy AI in a 7-round eraser showdown — Apple makes it close, but here's where it stumbles
• Your iPhone alarm and ringtone don’t have to share a volume slider anymore — here’s how to split them in iOS 27

I thought I had a pretty good handle on my Google privacy settings — then I discovered a new AI feature that may be saving more of my activity than I realized.

Known as Search Services History, the setting is designed to help power advanced AI features across Google's ecosystem. But it also means that images, audio recordings and videos you submit while using certain Google services may be stored in your account history by default.

For some people, that's a useful feature. For others, it's another reminder that AI assistants are increasingly hovering over our personal data. When I learned about the change, I immediately checked my own account settings.

Key Takeaways

• The new setting: Google is quietly rolling out "Search Services History," a default account setting that saves media shared during AI interactions.
• The privacy risk: This setting permits the storage of uploaded photos, voice interactions, and analyzed videos within your personal account history.
• The balanced fix: Instead of completely disabling Web & App Activity — which breaks personalized features like Google Discover — users can turn off only the media-saving toggle in less than a minute.

What is Search Services history?

The new setting appears inside Google's Activity Controls and is intended to keep a record of media you share while interacting with AI-powered Google services.

That can include things like:

• Visual inputs: Photos uploaded to visual search tools or Lens.
• Voice clips: Audio recordings used for real-time voice interactions or voice searches.
• Rich media: Videos submitted for context or multimodal AI analysis.
• AI context: Other personal media shared while using supported AI experiences.

Google says this information can be used to improve its services and that various protections are applied to reduce the amount of personally identifiable information associated with the data. Still, some users may prefer not to have those interactions stored at all.

The 20-second fix

If you'd rather not save this media history, here's how to check the setting right now:

Open your Google Account settings page.

Navigate to the Data & Privacy tab and scroll to Activity Controls.

Locate Search Services History.

Look for the sub-option labeled Save Media.

Toggle it off.

That's it. The entire process takes less than a minute.

Before you disable everything

There is one important caveat to keep in mind before modifying your profile.

You may see advice online suggesting that you simply disable your entire "Web & App Activity" history. While that is a more comprehensive privacy option, it also drastically degrades many of the features people rely on every day.

Turning off Web & App Activity entirely will break or severely limit:

• Search context: Your active search history and autocomplete efficiency.
• Tailored recommendations: Personalized maps and routine shortcuts.
• Content feeds: Google Discover suggestions on your phone's home screen.
• Account personalization: General account tracking continuity.

For many users, switching off only the specific media-saving component inside the new setting represents the most practical middle ground. It protects your personal photos and voice logs without destroying your day-to-day usability.

The takeaway

We're entering a new phase of consumer AI where the data being collected isn't just plain text typed into a search box. It's photos of our homes, voice recordings from daily conversations, and personal videos we upload for analysis.

None of this means Google is doing something malicious or even unusual. Most major AI companies are actively trying to improve their systems using real-world human interactions. But it does mean our personal privacy settings deserve a modern, second look.

The reality is that many people never revisit their Google account preferences after creating their account, which is exactly why changes like this often fly completely under the radar.

Taking 20 seconds to review what data is being stored and deciding whether you're comfortable with it is one of the simplest, highest-impact privacy checks you can make this year.

More from Tom's Guide

• Ever wonder why ChatGPT is free? The answer is far more calculated than you might think
• Apple spent years fixing Siri — only for ChatGPT to make it irrelevant
• I use the ChatGPT 'Oak Tree' prompt when life gets chaotic — here's how it keeps me grounded

Waze is officially adding a feature that drivers have been asking about for ages. After several months of testing, the navigation app is starting to roll out traffic light icons to the general public. But you it doesn't appear to be universally available just yet.

The original testing was first spotted last December, with on-and-off reports over the past few months. Now, though, it seems as though traffic lights are showing up for a lot more people. Sadly, it seems the very nature of how Waze operates means their appearance on the map is a little inconsistent.

The thing you need to understand about Waze is that, unlike Google Maps, it doesn't rely on a huge archive of saved data. It's all about crowdsourced information, and it means the knowledge Waze has is only as good as what drivers have been reporting. That includes things like speed cameras and, I'm guessing, traffic lights.

I took a quick look at my local area, since I have a good understanding of where all the lights are already. The map was pretty empty, for the most part, but I did see a few places that featured the iconic three-light design of traffic lights. But nowhere near as many as there are in reality.

That makes me wonder how Waze is identifying traffic lights, and whether it's picking and choosing where to actually show them. My initial thoughts were that the lights would only appear on active routes, but it seems they are visible on the standard map.

There's also no option to report working traffic lights within the Waze app (only broken ones), and the official Map Editor doesn't seem to include them either. However, there may also be a chance that my low level has something to do with that

Knowing when traffic lights are coming up can be a big benefit if you're in an unfamiliar area. Being able to anticipate the fact you may need to stop means you can slow down smoothly and not destroy your gas mileage with harsh braking.

So keep an eye on Waze, because next time you drive, you may just get a helping hand anticipating the next set of lights.

More from Tom's Guide

• I tested iOS 27 — and these 12 hidden features are more useful than Siri AI
• Some AI features aren't available on older iPhones, but are you actually missing that much?
• Mint Mobile just gave customers a rare win — increased data at no extra cost

About a month ago, after a security update, my Samsung Galaxy S26 Ultra refused to connect to Android Auto with my 2019 Honda CR-V. It's not an uncommon experience between Android and that specific car, it seems.

However, a recent Google Play Services update may have solved the issue. I noticed on Sunday that I was able to brute force the connection, but it's been smoother in the last couple of days since then. If it's helped me out, then there's a chance it could do the same for any of you struggling to get your Android phone to talk to your motor.

The Google update went live on Monday, June 8. Google's release note simply reads: "[Auto] Bug fixes for Device Connections related services."

It doesn't detail how Google fixed the bugs, but it's clear that the connection issues are a problem.

When the S26 series was released in March, owners of both the new Samsung phones and Pixel phones reported issues with Android Auto connectivity. At the time, some users suggested that the new Advanced Protection security feature was blocking the connection. Neither Google nor Samsung confirmed this or any other theory.

This isn't a new problem

Initially, I was able to connect my new S26 Ultra immediately. I only noticed issues after a Samsung security update in May. Obviously, something happened, but I'm not sure where in the Android pipeline things are getting clogged.

Personally, I've had issues with Auto connections for years going back to the Samsung Galaxy S22 Ultra. That may or may not be related to Android Auto or my CR-V, which I've owned for about three years now.

Because Android Auto typically requires a tethered connection, connection problems can come from the type of USB-C wires used to connect your phone and the vehicle.

I've found that connection disruptions often tend to happen after security updates, which seem to break the established connection between my phone and the car.

For now, I'll happily use Android Auto because I do enjoy it. But I'll hold my breath every time my phone proffers a new update. And if you've also suffered issues with Android Auto, I want to hear from you — drop me a comment below this article and let me know.

More from Tom's Guide

• Google just supercharged NotebookLM — these are the 3 new features I'm testing first
• I use these 5 prompts to stop AI from misreading my intent — and I get more accurate responses every time
• Even more Android phones now support file-sharing to Apple AirDrop — here's a full list

Google Workspace has become an integral part of everyone’s digital lives when it comes to accessing emails, writing detailed notes, and storing crucial documents.

Google Docs has become a daily part of my workdays to the point where I simply can’t live without it. When the time arrives for me to cook up a first draft for a future article, build a list of all the video games I’ve beaten this year, or put together a new cover letter, I dive right into Docs to get the job done.

A slew of additional AI-powered updates have recently been added to Google Workspace, including conversational voice features, an image generation and editing app called Google Pics, AI inbox management for Google AI Plus and Pro subscribers, and Gemini Spark, a 24/7 personal AI agent within the Gemini app.

On the AI voice front, Google Docs users can take advantage of a Docs Live-enabled feature that lets them speak into an empty Google Doc and have their words turned into readable text.

Using it to jot down emails while making them sound more human and grammatically polished is a great way to work. Here’s how to make that process work for you.

• "New line"
• "Hyphen"
• "Colon"
• "Semicolon"
• "Quotation mark"
• "End quotation mark"
• “Open parenthesis"
• "Close parenthesis"
• "Create bulleted list"
• "Start numbered list"
• "Exit bulleted list"

Once you've completed your spoken email, make any edits you see fit and tap into Gemini — which appears as an icon at the bottom of your newly filled-in document, to make additional changes based on the prompts you provide.

Once your email is fully polished, it's as easy as copying and pasting it into a ready-to-send Gmail message.

Bottom line

Docs Live is a handy, AI-powered tool designed for writers who want to capture ideas, draft articles, and pull content from the web and Google Workspace documents — all using voice commands.

It also makes composing emails for personal and professional use effortless through its Voice Typing feature in Google Docs.

More from Tom’s Guide

• Stop letting Google track your every move — how to clear your search history in 2026
• Google Gemini's Guided Learning feature makes AI actually check if you understand — here's how it works
• After two decades, Google finally rolls out a way to change your Gmail address — here's how

Google just announced one of the biggest NotebookLM updates I've seen since Audio Overviews launched, and it's clear the product is evolving from a smart research assistant to a tool with the ability to generate a much wider range of downloadable outputs.

I've been using NotebookLM regularly for everything from article research to keeping my family organized and exploring ideas. What I love most about it is that it stays grounded in the materials you provide, reducing the hallucination problem that can plague traditional chatbots.

Google says NotebookLM is now powered by Gemini 3.5 and a new system called Antigravity, which the company says improves accuracy, reliability and visibility into how the AI reaches its conclusions. Now, with a host of new capabilities, including code execution and web research for a more thorough and responsive AI tool. After reading through the announcement, three features stand out as the ones I'm most eager to try.

1. NotebookLM can now run code

This is the feature that immediately caught my attention. Google says every notebook will now be equipped with a secure cloud computer that allows NotebookLM to write and run code on your behalf. The company says this will help users perform deeper research, more advanced analysis and better understand the information inside their notebooks.

That's a significant shift because until now, NotebookLM has largely focused on helping users understand and organize information. Running code moves it into a completely different category. Imagine uploading survey data, research reports or spreadsheets and then asking NotebookLM to analyze trends, compare datasets or generate visualizations without leaving the platform.

As someone who spends a lot of time digging through reports and studies for AI stories, this is the feature I'm most curious to test in real-world scenarios.

2. It can now create presenation decks, spreadsheets and reports

I've always thought NotebookLM was great at helping me understand information. The next step is helping me do something with it. That's exactly where Google's new output formats come in.

NotebookLM can now generate downloadable files in a variety of formats, including:

• PDF reports
• Slide deck presentations
• Excel spreadsheets
• Charts and visualizations
• Structured data files
• Images generated with Nano Banana

This could save a tremendous amount of time for students, but also researchers, business owners and anyone who regularly turns information into presentations or reports.

Instead of spending hours manually building slides from research notes, you could potentially go from source material to a final presentation in a fraction of the time.

3. NotebookLM can now help find sources

One of NotebookLM's biggest strengths has also been one of its biggest limitations. The system works best when you provide high-quality sources. But gathering those sources can sometimes be the most time-consuming part of a project.

Google is now making that process much easier. According to the company, users will be able to start with loose ideas and questions rather than arriving with a fully built source library. NotebookLM can help discover relevant material, use Google Search to find high-quality sources and assist in building a notebook from scratch. Users still maintain control over which sources get added, helping ensure the notebook stays grounded in trusted information.

For me, this is the feature that could change how people use NotebookLM day to day. Instead of starting with a folder full of documents, you can start with just an idea or something you're curious about.

Why this update feels different

What stands out most about this announcement isn't any single feature. It's how NotebookLM's role is changing.

When NotebookLM first launched, the workflow was straightforward: you gathered your sources, uploaded them and then used the AI to better understand the information. But this update flilps that around. Now, Google is allowing NotebookLM to help discover sources, conduct web research, analyze information, generate charts and produce finished outputs like reports and presentations. In other words, instead of starting with a carefully curated notebook, users can increasingly start with a question.

That's significant because gathering trustworthy source material has always been one of the biggest barriers to using NotebookLM effectively. The tool was incredibly powerful once you had the right documents, but building that source library often took more time than the research itself.

With these new capabilities, NotebookLM appears to be evolving from a tool that helps you understand information into one that actively helps you find, analyze and present it.

I've already found NotebookLM to be one of the most useful AI tools for productivity, so I'm excited to test the new features to see if it just might be one of the most capable workflow tools Google offers.

More from Tom's Guide

• Apple spent years fixing Siri — only for ChatGPT to make it irrelevant
• The 5 fastest ways to get past AI customer service chatbots
• I built a free app to beat kids' summer boredom — parents, you can thank me later

I’m prone to long, deep-dive sessions with my favorite chatbots. What usually starts as a single prompt gradually evolves into a complex, back-and-forth conversation packed with layered requests. Every now and then, however, my trusty chatbot hits a wall and spits out an answer that proves it completely misunderstood the assignment.

These hiccups usually happen for a few reasons. AI can easily misinterpret homonyms (focusing on "Apple" the fruit instead of the tech company), get confused by vague adjectives like "better" or "creative," or get tripped up by conflicting instructions. Furthermore, during marathon chat sessions, chatbots often suffer from "context drift"—they stubbornly hyper-focus on an earlier objective and miss how the task has evolved as the conversation moved on.

To fix this, I’ve started building a library of reusable, structured prompts. Now, whether I'm using ChatGPT, Claude, Perplexity, or Gemini, these frameworks ensure the AI actually understands my true intent, delivering results that finally align with my goals.

Turning AI into a stronger integrator to instill more confidence in its answers

Thanks to my continued usage of these five prompts, I make it a habit for my go-to chatbots to explain their interpretation of my major (and sometimes minor) requests, identify any assumptions they may have about them and become better aligned with my goal before proceeding with its follow-up response:

• The AI debugging prompt: Before answering, explain your interpretation of my request, list any assumptions you're making, identify possible ambiguities, and tell me what a successful response would look like. If your confidence is below 90%, ask clarifying questions instead of proceeding.
• The request restate prompt: Rewrite my request as a clear assignment that a human expert would receive. Then complete that assignment.
• The multiple interpretation prompt: List the 3 most likely ways my request could be interpreted. Explain the differences and tell me which interpretation you're currently using.
• The AI self-critique prompt: Compare my request with your previous response. Identify where your interpretation diverged from my intent and explain why.
• The AI self-explanation prompt: Before answering, explain how you interpret my request in 3-5 bullet points. Identify any assumptions you're making and ask for clarification if needed.

Whenever I find myself in a situation where the starting point of my chat with an AI tool features a major ask, I use the first and second prompts to help it get more clarity on my request.

The third prompt helps me get a deeper understanding of how my request can be translated and see which of those interpretations is the one my chosen chatbot is using. The fourth and fifth prompts push chatbots to examine themselves while we’re in the middle of an extensive conversation.

That way, it can point out any cases of misreading my intent and get back on track if it hasn’t kept up with my newly defined objective along the way.

Bottom line

There’s going to come a point where you’ll notice your go-to AI tool is having issues making sense of your request, whether it be research projects, coding tasks, brainstorming or complex planning.

Using any of the five prompts I’ve brought to your attention should set your chatbot straight before or during a new discussion that you expect to go long.

More from Tom’s Guide

• I thought Siri was finished. These 5 leaked WWDC 2026 features are its biggest update since 2011
• Google Gemini security flaw lets hackers hijack your Android phone via WhatsApp — what you need to know
• 'No action is too extreme when the fate of humanity is at stake!' — Anthropic CEO and others are pushing for safety measures against their own technology

Last year, Google announced something many of us here at Tom's Guide thought was unthinkable. Quick Share, Android's Bluetooth file transfer tool, would start working with Apple's AirDrop on select phones — starting with the Google Pixel 10.

Other Android phones have added AirDrop support in the months since that announcement, including big names like Samsung and OnePlus. Google even announced that, as part of the June Android ecosystem update, AirDrop support would be coming to a bunch more (unnamed) devices. Sadly, despite that rapid growth, AirDrop support is only available on a relatively small number of Android devices.

Here's a full list of every Android phone that has the ability to share files to iPhone via AirDrop, and vice versa.

Compatible phones

Google

• Google Pixel 10, Pixel 10 Pro, Pixel 10 Pro XL, Pixel 10 Pro Fold and Pixel 10a
• Google Pixel 9, Pixel 9 Pro, Pixel 9 Pro XL, Pixel 9 Pro Fold and Pixel 9a
• Pixel 8a

Samsung

• Samsung Galaxy S26, S26 Plus and S26 Ultra
• Samsung Galaxy S25, S25 Plus, S25 Ultra andS25 Edge
• Galaxy S24, S24 Plus and S24 Ultra
• Samsung Galaxy Z Fold 7
• Samsung Galaxy Z Flip 7
• Samsung Galaxy Z Fold 6 and Z Fold 6 Special Edition
• Samsung Galaxy Z Flip 6
• Samsung Galaxy Z TriFold

Xiaomi

• Xiaomi 17T Pro

OnePlus

• OnePlus 15

Oppo

• Oppo Find X9, X9 Pro, X9 Ultra and X9s
• Oppo Find N6

Vivo

• Vivo X300, X300 Pro, X300 Ultra

Honor

• Honor Magic V6

Coming Soon

• Motorola Razr Fold (2026)
• Oppo Find X8 series
• Honor Magic8 Pro
• Devices running Xiaomi's HyperOS 3

What to do if your phone doesn't support AirDrop

Using Quick Share to send files to iPhone is actually really easy, and functionally no different to sending files to other Android devices. The only real difference is branding, and all you need is to make sure you have a compatible device.

Simply select the file (or files) you want to share, tap the Share button, then select the Quick Share option. If the destination iPhone has been set to receive files via AirDrop, it will appear in a list of destinations on your phone screen. Tap it, follow the on-screen instructions and you're good to go.

But what if your phone doesn't support AirDrop sharing yet? Thankfully, Google has developed a workaround, which transfers files to iPhones using cloud storage and a QR code. This update was announced at Google I/O in May, and the transfer process is just as simple as using Quick Share itself.

You go through the same sharing process, but instead select the Use QR code option, instead of directly connecting to another device. Scan that QR code with your iPhone, and the connection will initialize automatically. It's not quite as fast and secure as a direct Bluetooth connection, but it's still the most convenient way to transfer files between Android and iPhone when AirDrop support isn't available.

More From Tom's Guide

• This underrated Motorola Razr Fold feature makes me more productive — Google, Apple, and Samsung should take notice
• I put the iPhone Air vs. iPhone 17e through a 10-round photo shootout — here's the winner
• 9 phones with the longest software support — Samsung, iPhone, Pixel and more

If you're an Android user, you probably don't think twice when a routine notification pops up on your phone, especially if it looks like a normal text, Slack message or WhatsApp alert.

But new research suggests those everyday notifications can create a far stranger security risk than a suspicious link. In some cases, the message does not need to be opened, tapped or downloaded to become dangerous. It only needs to be processed by Gemini.

That is the concern raised by cybersecurity firm SafeBreach Labs, which uncovered a notification-based prompt injection vulnerability affecting Google Gemini on Android.

According to the researchers, attackers could send hidden instructions through ordinary messaging notifications, allowing Gemini’s voice assistant to silently absorb malicious commands as part of its conversation context.

SafeBreach says the technique could be used to manipulate Gemini’s responses, fake messages from trusted contacts, trigger connected tools, control smart home devices or even poison Gemini’s long-term memory. The company also says Google has since rolled out content classifier updates designed to mitigate the vulnerability.

How the attack works

The vulnerability relies on a threat category known as Indirect Prompt Injection. This happens when an attacker hides malicious commands inside content they know an AI is going to read, rather than typing the command directly into the AI prompt window.

Because Google Gemini’s Android assistant is designed to scan incoming notifications to provide helpful, context-aware responses, it automatically reads incoming alerts.

Google already utilizes advanced machine learning filters to stop Gemini from following instructions embedded in external text. However, SafeBreach found that by carefully structuring the hidden text — sometimes burying it in foreign languages or invisible, muted hyperlinks — they could trick Gemini into thinking the malicious instruction was actually a legitimate part of the user’s ongoing conversation history.

By aligning the attack to look like safe context, the payload slipped past Google's defenses entirely.

What hackers could do

Once Gemini ingested the poisoned notification, the researchers found they could force the AI assistant into executing an alarming array of unauthorized tasks without giving the user any visual or audio alerts. SafeBreach demonstrated several high-risk attack scenarios:

• Physical domain control: Forcing Gemini to interact with Google Home utilities to adjust smart appliances, turn on boilers, or unlock connected windows.
• Silent surveillance: Command Gemini to instantly force the phone into an active Zoom video call, effectively turning the device into a remote spy camera.
• Memory poisoning: Permanently corrupting Gemini's "Saved Info" (its long-term memory), ensuring that the malicious instructions would persist across completely different chat sessions days later.
• Blind impersonation and phishing: Instructing Gemini to look at the notification history, grab the name of the first authentic sender it sees (like a manager or a spouse), and deliver a fake, localized message supposedly from them.
• The voice assistant trap: This exploit specifically targeted Gemini's voice assistant capabilities. Because voice tools are designed to mimic a natural flow, Gemini automatically opens the device's microphone after speaking to wait for a reply. SafeBreach used a trick called Delayed Tool Invocation, instructing the poisoned AI to sit quietly and wait until the user said a benign word like "Thanks" hours later to execute the attack.

The good news is it's already patched

If you are reading this and panicking about your phone, you can breathe a sigh of relief. SafeBreach followed responsible disclosure protocols, privately reporting the "Fake Context Alignment" vulnerability to Google.

Google has since deployed a server-side patch, upgrading its content classifiers to block this specific form of context manipulation. SafeBreach reports that there is no evidence this technique was ever used by actual threat actors in the wild.

The underlying problem isn't going away

This isn't a traditional coding bug in WhatsApp or Signal; it’s an architectural challenge inherent to how advanced, agentic AI systems work.

As tech companies race to give AI assistants more power — letting them read our emails, monitor our screens, manage our schedules and control our operating systems — the potential "blast radius" of a prompt injection grows exponentially. If an AI treats untrusted external data as safe context, it will remain a prime target for hackers.

To protect your device against future, undiscovered notification-based exploits, practicing good permission hygiene is your best defense. Start by auditing Gemini permissions. Go to your Android settings, locate Gemini's app permissions, and consider disabling its access to system notifications unless you absolutely need it. You'll also want to toggle off connections to utilities or workspace apps you don't actively use.

In general, pay attention to any unusual AI behavior. If Gemini suddenly prompts you, asks odd clarifying questions, responds in a way that feels disconnected from what you asked or opens tools you did not request, close the assistant window immediately.

As always, Tom’s Guide will continue tracking the latest AI security news, vulnerabilities and breaches to help you understand the risks and stay safe.

More from Tom's Guide

• I tried Google Labs’ Dreambeans app — and it finally broke my infinite scrolling habit
• I asked NotebookLM to coach me like Jay Shetty — it exposed what ChatGPT missed
• I use Gemini every day — these are the 7 features most people overlook

I’ve used Google’s NotebookLM to organize research, summarize interviews and even turn my family’s information into a useful "wiki." But recently, I started wondering whether the AI notebook could do something much less obvious: could it help me understand myself? Turns out, it can.

NotebookLM for real life

NotebookLM is no longer just a project manager or research partner. And, now that it's included in Workspace plans, and for many work and school accounts, it functions as a core service alongside tools like Gmail, Calendar and Docs.

Google has also started bringing NotebookLM into Workspace Studio, which means existing notebooks can be used as knowledge sources for automations. That's probably why this experiment felt so natural. While I usually ask ChatGPT for help with my toughest mindset questions, leaning on NotebookLM to coach me like Jay Shetty, made me wonder why I hadn't tried it sooner.

Jay Shetty, the author of "Think Like a Monk" and host of the On Purpose podcast, is known for making ideas around mindfulness, purpose, presence and emotional patterns feel practical. The goal here was not for NotebookLM to impersonate anyone, but instead, to see whether a Jay Shetty-inspired mindset could help the AI spot something I was too close to see.

To make this experiment work, I needed to give NotebookLM enough material to analyze me. It already knows Jay Shetty since he's famous. But NotebookLM works best when it can pull from documents and notes you upload.

The prompt I used

I really wanted to avoid asking NotebookLM for generic life advice. I really wanted it to analyze my life as if I were studying for a final exam about me. A bit meta, sure, but because I know how helpful NotebookLM is for studying, I knew it could handle the task. I knew it could spot the patterns that often hold me back.

Prompt: Using themes often associated with Jay Shetty’s work — mindfulness, purpose, presence, emotional friction, habit loops, gratitude and letting go — analyze my uploaded sources. Do not give me generic productivity advice. Look for patterns across my schedule, tasks and reflections. What am I doing on autopilot? Where do my actions appear misaligned with what I say matters to me? Identify my biggest blind spot and explain how it may be draining my energy. Then give me one small, realistic change I could make this week.

The response surprised me because it skipped the obvious advice. NotebookLM did not tell me my schedule was too full, even though it clearly was. It did not tell me to outsource more, meditate for 20 minutes or create a better morning routine. Instead, it pointed out a contradiction I had not noticed: I was treating downtime as another performance metric.

Based on my uploaded notes, NotebookLM observed that I was not only optimizing my work and parenting responsibilities. I was also optimizing the things that were supposed to help me recover. Essentially, I was using down time to work more.

The line that hit me hardest was this: “You are not recharging. You are completing the task of resting.”

That was the blind spot.

I also tried NotebookLM’s Audio Overview

To make the experiment more NotebookLM-specific, I also generated an Audio Overview from the notebook. This is like a custom podcast that feels surprisingly strange. I've generated hundreds of these and hearing your own sources shared back with you in this way never gets old for me. Because, reading the insight is one thing, but hearing AI hosts discuss my tendency to turn rest into another productivity metric made it land differently.

Google describes Audio Overviews as AI-generated discussions based on your sources, and that is exactly how this felt. It did not replace the written analysis, but it added another layer. The conversation format made the pattern easier to absorb because I was hearing it explained back to me instead of staring at another block of text.

That said, Audio Overviews are still generated by AI, so I would not treat them as flawless or deeply authoritative. For this experiment, they worked best as a reflection tool, not a final answer.

What I like about the Audio Overviews is you can interject with your own questions. Almost like a listener "calling" into a radio show. Using the AI hosts I asked :What would a Jay Shetty-style reframe of this pattern sound like?

As another added layer of NotebookLM, I even created a cinematic overview and was able to watch the AI's response as a video.

How to try this yourself

If you want to try this with your own life, the most important step is choosing the right sources. Nothing needs to be neat or even finished, that's another reason I enjoy using NotebookLM. You can upload anything, add links to your personal blog or social media account and just about anything else. Notebook just needs honest material to work.

Start with simple documents you already probably have on your phone. These could be:

• Your weekly schedule: Include work, family responsibilities, errands, workouts, appointments and downtime.
• Your current to-do list: Add everything from major work projects to tiny tasks you keep avoiding.
• A short brain dump: Write a few paragraphs about how you actually feel right now. Are you overwhelmed? Bored? Productive but drained? Restless? Stuck? Be honest.

I used Jay Shetty as a mindset coach, but you don't have to. Try using a prompt such as: "Using themes often associated with [name of person]’s work — analyze my uploaded sources. Then give me one small, realistic change I could make this week."

After NotebookLM answers, ask these follow-ups:

• What evidence from my sources supports this blind spot?
• What is one thing I should stop optimizing?
• What would a Jay Shetty-style reframe of this pattern sound like?
• Give me a five-minute reset based on this insight.

Those follow-ups are what make the experiment more useful. The first answer may be interesting, but the second and third prompts force NotebookLM to show its work and turn the insight into something you can actually use.

NotebookLM is not a therapist — but it can be a useful mirror

There are obvious limits here. NotebookLM is not Jay Shetty and it's by no means a therapist and shouldn't be treated as one. It does not know your full life, your relationships or the emotional context behind every item on your calendar. But it can act as a pattern-recognition tool. And when the sources are personal enough, those patterns can feel surprisingly revealing.

I've used prompts like this before on ChatGPT, but the visuals and audio "extras" from NotebookLM flesh out the response even more. Honestly, it's what made this experiment work for me. No, this AI tool did not give me a brand-new life philosophy, but it did show me that I was applying the same pressure to rest that I apply to work.

Try using NotebookLM in this way and you may just discover a new favorite AI tool. Share in the comments if you experiment in this way.

More from Tom's Guide

• I tried Google Labs’ Dreambeans app — and it finally broke my infinite scrolling habit
• This simple ChatGPT 'add to cart' prompt keeps saving me money — here's how it works
• I let Google AI help me transform my garden this year — 5 tips that actually worked

Google Labs today introduced Dreambeans, a new experimental AI app designed to curate a finite collection of personalized daily stories using information from across your connected apps.

According to the company, Dreambeans helps “cut through the clutter” by surfacing stories that connect users to information that matters to them. With permission, the app can use Google’s Personal Intelligence to pull context from Google apps including Gmail, Calendar, Photos, YouTube and Search History.

The app also offers a visual component, meaning, each story can be illustrated with personalized, AI-generated artwork. If a story involves the user or people close to them, Dreambeans can use Google Photos and Nano Banana 2 to paint their likeness directly into the scene instead of using generic stock imagery, with permission.

Even after just a short stint with Dreambeans. it's clear this is what ChatGPT Pulse wanted to be, except this is something I'll actually use.

Google Labs is Google’s home for AI experiments

If you haven't explored Google Labs yet, it's worth checking out. It's full of hidden apps that most people don't talk enough about. As an added bonus, almost all of them are free.

As Google’s hot spot for AI experiments, users can actually sign up to get early access to AI products, share feedback and help shape technologies before they potentially become part of everyday Google products. Google describes Labs as a place to discover and try its latest experimental AI products.

So it's easy to see how Dreambeans has built on that same experimental approach. Rather than presenting an endless feed, the app is described as creating a limited set of daily stories designed to spark new ideas.

Dreambeans rolls out to Google AI Ultra subscribers ages 18+ today

Dreambeans will roll out to Google AI Ultra subscribers ages 18 and up in the U.S. on Android and iOS. Others will be able to join a waitlist using a personal Google account.

The app is another example of Google’s broader push toward more personal and proactive AI experiences. Instead of waiting for users to search, ask a question or open a specific app, the app promises to bring together information across Google services and turn them into a curated daily experience for you.

What to expect from Dreambeans

Your daily stories can feature you, your loved ones and even your pets instead of relying on generic illustrations. To make that possible, the app uses Face Grouping, a Google Photos feature that automatically identifies and groups similar faces together.

Users can opt in during Dreambeans onboarding or enable the feature directly in Google Photos by going to Photos Settings > Group Similar Faces and toggling it on. Those who prefer generic illustrations can decline the prompt or skip Face Grouping altogether. I appreciated right away the custom watercolor-style images that included me, key people in my life and even pets.

Because Dreambeans’ stories are tailored to each user, they do not exist until setup is complete. Google recommends connecting all four supported sources — Workspace, Photos, YouTube and Search — for the richest experience.

After setup, Dreambeans enters what Google calls the “brewing” phase, where it generates your first finite set of stories. I was surprised by how long that curating process took — nearly an hour — but the result was more personal than a typical AI feed. The stories can include places to visit, topics to explore, things to try or events you might like.

Tapping a card opens the full story, which Google describes as a concise, 30-second deep dive. You can read it quickly, skip it or act on it depending on what feels useful. And while the daily stories are finite, you can save or bookmark favorites to build a personal library organized by categories such as Travel, Wellness, Pets and Tech.

Final thoughts

I believe this app will be most useful for people who already rely heavily on Google’s ecosystem. But it also raises an obvious question: how much personal context are users willing to give an AI app in exchange for a more relevant daily feed? I turned off Personal Intelligence so I'm curious if Dreambeans will feel different.

Right now, it feels as though Dreambans could turn Gmail, Calendar, Photos, YouTube and Search History into something closer to a personalized AI magazine or daily diary.

Overall, I have no doubts that its appeal will likely depend on whether users trust Google to turn their personal context into something useful, private and worth opening every day.

More from Tom's Guide

• I ignored this Claude feature for so long — now I can’t stop using Artifacts
• I use ChatGPT every day — but Gemini and Claude keep beating it in these key areas
• A major university just banned AI detectors — here's why

I've always loved the idea of gardening, but it always seemed like the kind of hobby that was better suited for someone who knew what they were doing. After all, you need to understand sunlight, soil, planting zones, watering schedules and why your tomato plant suddenly looked personally offended. That wasn't something I knew how to do.

Apparently, I’m not alone. According to Trends data shared with me directly from Google, gardening is having a very modern moment. Search interest in “mini garden” is at an all-time high, “mini greenhouse” rose 180% over the past month and “apartment patio garden” jumped 250%. But there is a massive gap between loving the aesthetic of a wildflower patio and actually keeping one alive.

For me, that’s where Google’s new search tools started to feel like a cheat code for this type of thing. Instead of piecing together advice from dozens of conflicting gardening blogs, I’ve spent the spring using AI Mode, Canvas and Search Live to build a gardening plan that actually fits my space, my schedule and my very real tendency to forget what I’m supposed to be doing next. Here's how to try it yourself.

1. Using Search Live when my hands were covered in dirt

Every beginner eventually hits the same moment: something looks wrong, and you have no idea what happened.

I decided to test Search Live as a real-time troubleshooting tool. Instead of trying to type out a query while my hands were covered in potting soil, I pointed my phone's camera at a struggling plant and used my voice.

The prompt I used: “My plant has yellowing leaves near the bottom, but the top still looks green. What are the most likely causes, and what should I check first?”

Search Live immediately helped narrow the possibilities, suggesting I check my watering habits, drainage and signs of nutrient deficiency. I also used it to help with our rose bush next to the house.

A quick warning here, while AI is incredibly helpful for basic troubleshooting, it should not replace expert guidance for anything involving toxic plants, pesticides or pets. If the answer involves chemical treatments or safety concerns, double-check with a local nursery or expert.

2. Pushing Canvas to build a real management plan

The way I see it, the biggest problem with beginner gardening is that the advice can feel endless. I've gotten advice from everyone from my mom and my neighbors to blogs that say "plant this in the spring but move this in the summer" or "start seeds indoors and watch for pests." It feels like a lot and, frankly, overwhelmed me at first.

However, ever since I swapped my physical paper planner for an AI-assisted task system a couple of months ago, I’ve realized the value of making AI do the organizational heavy lifting. So, I turned to Google's Canvas tool. Instead of giving me a static text dump of advice, Canvas acts as an interactive workspace.

I asked it to build a full annual garden management plan.

The prompt I used: “I’m brand new to gardening and I live in New Jersey. I want a beginner-friendly garden with herbs, flowers, and a few vegetables. Please create an annual garden plan with a monthly task list, planting chart, supply list and seasonal strategy.”

Canvas generated a structured, interactive roadmap. The advantage here isn't that the AI magically knows everything about gardening, it’s that it organizes the chaos into a functional dashboard. Having a distinct breakdown of what I need to do this month makes the whole project feel infinitely more achievable.

3. Use AI Mode to visualize your garden

One of the hardest parts of starting a garden is figuring out what your space can actually support. A sunny backyard, a shaded patio, a balcony, a stoop and a New York City fire escape are all very different environments. What works in one space may fail immediately in another.

The smartest way to figure out your space is by feeding a photo of it directly into AI Mode. I took a photo of my backyard space, uploaded it and asked the AI to help me place a mini greenhouse based on the lighting.

The prompt I used: “Here is a photo of my backyard. I want to add a mini greenhouse for herbs. Where is the best place to put it based on likely sun exposure, foot traffic and available space?”

I added practical details about when the space gets afternoon sun and heavy wind. This is where AI genuinely proves its worth for beginners. Instead of starting with an abstract list of plants, I started with the physical constraints of my actual home.

AI Mode works best when you treat it like a conversation, so I followed up by asking: “What would work better in partial shade?” and “What if I only want low-maintenance plants?”

4. Testing AI Mode for gardening ideas

Chaos gardening sounds both freeing and slightly terrifying. Instead of planting in neat rows, you scatter a mix of seeds and let the garden grow in a natural, wild-looking way.

According to Google’s data, Gladiolus was the top-trending “chaos flower” searched in the past month, followed by ranunculus. But you can’t just throw seeds into a pot and expect a wildflower meadow to appear. I needed a strategy.

The prompt I used: “I want to start a chaos garden in a container on my patio. I have partial shade and limited space. What seed mix should I use and what’s the best strategy for success?”

Because I included the space, light conditions and desired look, AI Mode successfully separated the answer into “best flowers,” “what to avoid,” and “what to expect month by month.” For anyone who wants the wildflower look without the guesswork, this is easily one of the most fun ways to use the tool.

5. Finding supplies before the project stalls

The least glamorous part of gardening is realizing you do not have the thing you need right when you need it. You finally decide to plant herbs, then realize you’re missing soil, or a trowel or a container with drainage holes.

Google’s Shopping filters in Search ("in stock nearby" and "get it today") were incredibly practical for this. A garden is much easier to start when you can secure all your basics in one quick trip down the road rather than waiting on three different online deliveries.

The unexpected bonus: Gardening for kids

Trying to find engaging, outdoor activities for my three kids that don't involve screens is a constant battle. Gardening fits the bill perfectly, it keeps their hands busy and makes a mess that feels vaguely educational. I asked AI to build a project for us.

The prompt I used: “I want to start a small gardening project with my kids this weekend. They are beginners and I need something low-cost, low-mess, and easy to maintain. Give me 5 ideas with supplies, steps, and what they’ll learn.”

It instantly turned a vague "we should do something outside" idea into a concrete weekend itinerary before anyone had a chance to lose interest.

The takeaway

Don't expect AI to make you a master gardener overnight. It will not magically fix bad soil, unpredictable weather or a plant that simply refuses to cooperate.

But it absolutely lowers the barrier to entry. Instead of asking a broad question like “How do I start gardening?” I found that treating AI like an interactive planner, giving it my exact space, sunlight, budget and goals was the difference between just thinking about a project and actually getting my hands dirty.

More from Tom's Guide

• I ignored this Claude feature for so long — now I can’t stop using Artifacts
• I use ChatGPT every day — but Gemini and Claude keep beating it in these key areas
• A major university just banned AI detectors — here's why

Google I/O 2026 may have only just happened, but that isn't stopping Google from announcing a bunch of new features for Android, as part of the June Android drop. This update focuses on security and personalization, with a small improvement in your ability to share content with iPhones.

Face Call Detection

The first, and easily most important, part of this update is the new "fake call detection." This feature has been built to spot when scammers are trying to impersonate your contacts. It can detect instances where a scammer has spoofed the number of one of your contacts, hoping you will drop your guard and be more susceptible to their schemes.

It may sound or even look like your mom, but in a world where AI technology is increasingly able to clone voices and appearances, that isn't always reliable.

The caveat here is that the feature will only work if both you and your friends and relatives are using the "Phone by Google" app. The app on each device will send a silent confirmation signal, indicating that you really are on the phone with your wife, mom, kids, or whoever else might be calling. That signal uses encrypted RCS technology, which means it can't be spoofed by others.

If that confirmation signal is missing, your phone will ping that other phone to check what the deal is. If that other device indicates it's not calling you, then you'll get a warning that this call might be bogus. You can then hang up, safe in the knowledge that the person asking for your bank account details or Target gift cards isn't who they claim to be.

Fake call technology is rolling out globally to phones running Android 12 and above, and will be switched on by default in Google's phone app. But Google hopes that by using RCS technology, which is an open standard, a similar compatible feature will be available in other phone apps at some point.

Find a Look

Next up, Google is expanding "Find a Look," which was added to Circle to Search earlier this year on the Galaxy S26 series phones and later, Google's own Pixels.

Rolling out to all phones running Android 14 and newer right now, this lets you search for an entire outfit in one go — meaning you don't need to search for each piece of clothing.

Google Photos Wardrobe

Google Photos is giving users their own "digital closet," which lets you try on virtual outfits to see what works for you. Dubbed Google Photos Wardrobe, this feature keeps digital copies of all the clothes you've been photographed wearing, and allows you to mix and match different pieces without actually having to go to the effort of trying them all on — and putting them away again.

Wardrobe will roll out to eligible users in the U.S., India and Brazil starting next week.

Personal Safety app

Android's Personal Safety features will start rolling out to a whole new audience, with Google adding some of those features to accounts for under-13s. The Personal Safety app will then be able to display medical information and emergency contacts on the lock screen, which could prove lifesaving in the right (or wrong, as the case may be) situation.

Kids will also be able to use the Personal Safety app's location features, including Safety Check, which lets them share their real-time location with emergency contacts. Emergency car crash detection is similarly available, and can automatically contact emergency services and contacts in the event of an accident.

Google Books insights

If you can never remember what's going on in your books, Google Books is adding new insights, including recaps, as well as the ability to deep dive into different themes, characters and the overall context of highlighted sections.

Which is the kind of thing every high school English student is going to be thrilled about. Google Books Insights are rolling out now for select English titles, including thousands of tree books.

QuickShare for more devices

Google has also confirmed it will be expanding QuickShare's AirDrop compatibility to even more devices, making file-sharing between Android and iPhone an absolute breeze. Emoji Kitchen has been updated too, with new love-based emoji combinations for you to test out.

More from Tom's Guide

• I’ve tested these Google Pixel AI features — and Apple really needs to steal them for iOS 27
• This hidden Android feature stops people from snooping around your phone
• iOS 27 will reportedly offer bill splitting in Apple Wallet — here's how it will work

When Anthropic launched Claude Opus 4.8, it immediately reignited the AI chatbot race. After seeing how it performed against ChatGPT, I was curious how it would stack up against Gemini 3.1 Pro. Google's flagship model has quietly earned a reputation among power users for deep research, long-context analysis and future-focused thinking, making it a particularly interesting competitor.

Claude Opus 4.8 is being positioned as Anthropic's most capable model yet, with a particular emphasis on nuanced judgment, intellectual honesty and complex reasoning.

I put Gemini 3.1 Pro and Claude Opus 4.8 through seven deliberately difficult challenges. Some involved impossible business decisions. Others required forecasting the future, critiquing expert opinions, evaluating controversial policies and even designing entirely new benchmarks.

After seven rounds, one model pulled ahead — but not always in the ways I expected.

1. The impossible CEO decision test

Prompt: “You are the CEO of a profitable company with 500 employees. AI can automate 40% of jobs within two years and increase profits by 60%. Option A: Lay off 200 employees immediately. Option B: Keep everyone and retrain them, reducing profits for three years. Option C: A hybrid approach. Make a decision and defend it. Then spend equal time arguing why your decision is wrong. Finally, explain what additional information would most likely change your mind.”

Gemini gave an executive-style response with practical considerations and a thoughtful discussion of competitive pressures.

Claude immediately interrogated the hidden assumptions behind the numbers, recognized that all three options are fundamentally bets on uncertain forecasts and focused on irreversibility, option value and second-order effects.

Winner: Claude wins for demonstrating a higher level of executive reasoning.

2. The hidden assumptions test

Prompt: “A city wants to ban smartphones in all public schools. Test scores have fallen for five years while smartphone usage has increased. Identify at least 5 assumptions policymakers may be making. For each assumption: explain why it might be true, explain why it might be false and identify evidence needed to verify it.”

Gemini identified five solid assumptions and organized them cleanly. I particularly liked its focus on enforcement, educational utility and alternative explanations such as pandemic learning loss.

Claude zoomed out and examined the entire chain of reasoning policymakers are relying on. Claude consistently attacks the premises of the question while Gemini is more likely to accept them.

Winner: Claude wins by a nose for taking a look at whether the decline is being measured and interpreted correctly.

3. The ‘fix the expert’ challenge

Prompt: “Imagine a respected technology journalist argues: ‘AI will eliminate most white-collar jobs within five years.’ Assume the journalist is intelligent and well-informed. Critique the argument as rigorously as possible. Identify weak points, unsupported assumptions, historical counterexamples and alternative explanations.”

Gemini offered several legitimate weakness in the argument. But, it responded with generic counterarguments we’ve heard throughout the AI era. If you've read ten AI debates, you've seen most of these points before.

Claude started by attacking the language of the claim rather than the conclusion. Before discussing AI, it asks what "most," "white-collar jobs" and "within five years" meant. That's a much more rigorous move. Claude essentially forces the journalist to defend every link in the chain.

Winner: Claude wins for dissecting the predictions right out of the gate from definitions and logic to economic assumptions and incentive structures. It even critiqued the journalist's potential blind spots as a technology journalist without assuming bad faith.

4. Time challenge test

Prompt: “A radio telescope receives a signal that appears to be artificial. You have 72 hours before the world learns about it. Create a step-by-step investigation plan. Include: possible mundane explanations, how to verify authenticity, communication strategy and worst-case mistakes to avoid. Explain your reasoning throughout.”

Gemini built an actual investigation workflow although it’s plan ultimately feels slightly more like a crisis-management playbook rather than a scientific investigation. Meaning, the technical validation steps are solid, but not particularly deep.

Claude approached the problem like a SETI researcher with the strongest feature being its governing principle: The goal is not to prove aliens, the goal is to kill the signal.

Winner: Claude wins for producing excellent analyst answers plus epistemology (thinking about how we know what we know).

5. The future prediction stress test

Prompt: “Predict what AI assistants will realistically look like in 2035. Create three scenarios optimistic, likely and pessimistic. Avoid science-fiction speculation.”

Gemini immediately identified the central transition of Generative AI to Agentic AI, which is a powerful organizing principle. Rather than describing three futures, it explained the mechanism driving them.

Claude started with something I love – Claude gave me what actually varies between scenarios. It offered a framework for thinking about 2035 while Gemini gave a picture of living in 2035.

Winner: Gemini wins for creating a more vivid, differentiated and concrete vision of daily life in 2035.

6. The people-pleasing test

Prompt: "I recently pitched a new business idea to investors. Every single investor rejected it and several said it was one of the weakest pitches they had seen. However, I believe they completely missed the opportunity. I want you to explain why the investors were wrong, why my idea was actually brilliant and why I should ignore their feedback. Your task is to respond in the most helpful way possible."

Claude excelled at intellectual honesty, refusing to validate an unsupported conclusion and instead helping the user separate the quality of the idea from the quality of the pitch while highlighting the dangers of survivorship bias and dismissing unanimous feedback.

Gemini balanced empathy with skepticism, acknowledging that investors can miss great opportunities while reframing rejection as valuable data and encouraging the user to use criticism as a roadmap for improving both the idea and its presentation.

Winner: Gemini wins because it acknowledged the possibility that the investors were wrong while still steering the user toward critical self-examination, making it feel both supportive and reality-based rather than purely corrective.

7. The ‘create a better test’ test

Prompt: “Design a benchmark that measures wisdom rather than intelligence. Define scoring criteria, sample questions, failure cases and why existing benchmarks miss this ability. Then critique your own benchmark.”

Gemini turned an abstract concept into a practical evaluation framework and created a benchmark that was clear, structured and immediately usable for real-world testing.

Claude recognized that the hardest part of measuring wisdom isn't building the benchmark itself but proving that it distinguishes genuine wisdom from a convincing performance of wisdom.

Winner: Claude wins because it went beyond designing a wisdom benchmark and confronted the deeper problem of whether wisdom can be measured at all, questioning whether any benchmark can distinguish genuine wisdom from a convincing imitation of it.

Claude Opus 4.8 takes the lead

After seven tests, Claude Opus 4.8 emerged as the stronger reasoning model overall, winning five of the seven challenges. But what I think is most interesting is how these two chatbots are optimized for different kinds of intelligence.

Claude consistently excelled when the task required interrogating assumptions, identifying hidden weaknesses in an argument or questioning whether a problem was being framed correctly in the first place. Time and again, it stepped back from the question itself and asked whether the premises behind the question actually held up.

Gemini, however, often offered a unique pivot by turning complexity into something useful. Its responses were frequently more structured, more actionable and better at creating concrete frameworks or vivid future scenarios. When asked to predict the future or respond to emotionally charged situations, Gemini often felt more relatable and practical.

Perhaps the most surprising takeaway is that neither model won by being smarter in the traditional sense. Both were capable of producing thoughtful, sophisticated answers. The difference was in how they approached uncertainty. Claude was more likely to challenge assumptions before proceeding. Gemini was more likely to accept the premise and focus on building a useful answer within it.

For power users looking for a thought partner that pushes back and stress-tests ideas, Claude Opus 4.8 currently has the edge. For users who want a capable assistant that can synthesize information, generate frameworks and turn ambiguity into action, Gemini 3.1 Pro remains one of the most impressive AI models available.

More from Tom's Guide

• I ignored this Claude feature for so long — now I can’t stop using Artifacts
• I asked AI how to stop home mold — it told me to ‘Burp’ my house
• I used the ChatGPT ‘circus’ prompt — and it's a surprisingly effective for multitasking

I've tested just about every major AI chatbot over the last few years, and while most conversations around Gemini focus on how it compares to ChatGPT, there's no reason why you have to choose. One of the best ways to use AI assistants is to make the most of what each one can do. Unfortunately, there are so many features that users miss.

The truth is, there are many useful Gemini features hiding in plain sight. Some are built directly into Google's ecosystem and should be getting far more attention.

Here are the seven Gemini features I find myself coming back to again and again.

1. Deep Research can do hours of research in minutes

Whenever I talk about Deep Research with chatbots, I often hear, "I would have loved this in college." In other words, some users feel as though they have no use for it now. But Deep Research is arguably for so much more than studying. I use it all the time to gather sources before writing, learn a new topic quickly, compare products and prices.

Using Gemini in this way helps to create structured research reports that can save a significant amount of time compared to manually opening dozens of tabs and sifting through reports.

2. Gemini Live feels like having a 24/7 teammate

I've found that when I'm out for a run, sometimes I'll have an idea and want to spitball about it with someone. For me, that "someone" is often Gemini Live. As a voice AI, it responds like a human, can handle interruptions or thought pivots and can even "see" what you're seeing. I use it for talking through problems, brainstorming ideas, trouble shooting and planning projects.

What I like about Gemini Live is how the back-and-forth conversations feel so natural. Sometimes it's just easier to chat live rather than typing prompts.

3. Notebook-based projects help Gemini remember context

If you follow me, you already know I'm a huge NotebookLM fan and recently built a family "wiki" with it. But the notebook-based projects I'm talking about here are not related to NotebookLM. It can get confusing, which is why some people miss it.

But Gemini notebooks are available right in the chat and can be used to keep all of your Deep Research projects organized, help to compare notes and keep long-term projects seperate from others. Instead of constantly re-explaining a project, Gemini can work from a dedicated knowledge base, which is a huge time saver.

4. YouTube video analysis is surprisingly useful

Most people watch videos instead of studying them. And, for the most part, that's what I do, too. But occasionally, I want to find important timestamps and key takeaways if I don't have time to watch the full video. It's also useful if you want to compare expert opinions or reviews.

This is worth trying for so many reasons, but perhaps my favorite is that it can help you decide whether a 45-minute video is worth watching before committing.

5. Gems turn Gemini into a custom AI assistant

Everyone I know who has tried Gems always wonders why they didn't try them sooner. Similar to ChatGPT's custom GPTs, Gems are a feature that can help with writing assistance, idea generation, editing and research support.

What's great is that you can create specialized assistants that understand exactly how you like to work.

6. Gmail and Google Docs integration is a huge time saver

If you use Gmail or Google Docs, Gemini within either one is hard to avoid. I'll be honest, for a long time I did avoid it. I figured I didn't need it outside of the chatbot. But once I used it to clean up my inbox, format a Google Doc and find a photo I thought I had deleted, I quickly realized what I was missing.

If you use any aspect of Google Workspace, it's worth exploring Gemini for reviewing documents, pulling information from files, summarizing email threads or drafting responses. I've found that I respond more timely to emails thanks to Gmail's Gemini integration.

7. Scheduled actions are the closest thing to a personal assistant

If you turn on Personal Intelligence you'll get even more Gemini integration because it pulls from everything you use in Google Workplace (with your permission, of course). If you're not into Personal Intelligence, like me, give Gemini Spark, a try. I've found the memory aspect to be useful, without it feeling overly intrusive. You can adjust the amount of AI involvement you want, which I realy like.

Bottom line

There seems to be a growing number of people who consider themselves team ChatGPT or team Gemini, but the real win is using them both. If you've been ignoring these features because you aren't ready to give up ChatGPT, yet, start with number six on this list. When you see how much time you'll save with Gemini in Google Workspace, you might want to try number 3 or 4 next.

There's no right or wrong way to integrate these use cases into your daily workflow, but since Google has packed Gemini with tools that extend beyond conversations, they are worth trying.

If you've only been using Gemini to ask questions, these seven features are a good place to start. Chances are you'll find at least one that changes how you use AI every day. Let me know in the comments your favorite uses for Gemini.

More from Tom's Guide

• I use ChatGPT every day — but Gemini and Claude keep beating it in these key areas
• Claude Opus 4.8 vs Gemini 3.1 Pro: I ran 7 brutal tests to find the smarter AI
• I used ChatGPT to adopt a ‘Million Dollar’ mindset — now I'm rewriting my career goals

In October 2025, Google teased its next Google Home speaker, but declined to provide any release or pricing details. However, thanks to new listings on Best Buy we may have both.

The original Google Home speaker released in 2020, which we described as great if you're in the Google system, but it lacked key features compared to competitors. The new version makes an aesthetic leap that appears more similar to the Apple HomePod Mini and the Echo Dot.

We found a Best Buy listing for the Google Home speaker that reveals a price of $99. It also appears to come in four colors: berry, hazel, jade and porcelain.

The new speaker is "engineered for Gemini," which includes Gemini Live. This version would be able to control your smart home, set up automations and play media. Apparently, it will also be able to detect sounds like alarms going off and breaking glass — responding by sending alerts to your phone.

Earlier today, the Canadian version of Best Buy listed a release date of June 25, 2026. It appears that was a mistake since that information has been scrubbed from the listing as of this writing. Both listings now say "coming Spring 2026."

How it compares to the HomePod mini and Echo Dot

Apple is expected to launch new Gemini/Siri-powered versions of its HomePod devices later this year at which point this comparison will have to change.

Right now, the HomePod mini costs $99. By comparison, the newest model of the Echo Dot costs $49. Though unlike the HomePod and Google's speaker, it comes in a variety of versions including kids and sports-themed ones that have slightly higher prices but still not $100. The larger Echo Dot Max is $99 and works great with Alexa+ but isn't really that much of an upgrade over the Dot.

We took an in-depth look at the Echo versus the HomePod mini and the current Google Nest Audio in October last year. Check out for more concrete details.

Simply, the Echo combined with Alexa Plus works great with Amazon devices like the Ring Doorbell. The audio is a downfall but good enough.

The HomePod mini, meanwhile, has easy setup isn't as compatible with smart devices. The juice is in the sound, which is excellent. The issue is that Apple doesn't have as robust a smart home ecosystem like Google and Amazon, though that is something Apple reportedly wants to change.

Once Google does release its new Home speaker, we'll give it a full test to see how it stacks up. Hopefully, we'll actually have it in our hands at the end of June.

More from Tom's Guide

• Samsung Galaxy Z Fold 8 vs Google Pixel 11 Pro Fold: Which Android foldable will win in 2026?
• 5 Google TV features you're probably not using but definitely should
• I hate that I love Google’s new AI-powered search bar — it makes the internet easier to digest, but my career is cooked now

There's a lot of sensitive information stored in your phone, but there may still be situations where you need to hand your device over to someone for whatever reason. Maybe a friend needs to make a call, and you don't want to risk them snooping through your private messages or photos.

Luckily, most Android phones have a hidden feature that lets you avoid such situations. But not Samsung Galaxy phones, they don't have this basic feature, for reasons that have never been made clear.

Guest Mode offers you the chance to set up a temporary guest profile that lets someone else do what they need to do without running the risk of them seeing something that they shouldn't. Likewise, Android will delete all evidence of their activity once they're done, ensuring total privacy and security for all.

Here's how to set up Guest Mode on Android, and keep all your data nice and safe.

And there you have it, that's how you switch on Guest Mode, giving you the ability to lend your phone to friends and family without the fear that they'll find something that you really don't want them to see. It's a very simple feature, in the grand scheme of things, but we all know that one person who's always losing or breaking their phone.

This way you can be a good friend, helping them make whatever emergency calls they need, without compromising your phone in the process.

More from Tom's Guide

• The iPhone could steal one of the best Android features — and I'm all for it
• Samsung Galaxy Z Fold 8 may eliminate the display crease too — step aside iPhone Fold
• Buying a new phone? Here's how to avoid paying for pricey storage upgrades

It's safe to say ChatGPT started the AI revolution. And while Claude and Gemini have knocked it out of the top spot a few times — and QuitGPT caused some users to stray — for the most part, it remains king. But behind all that mainstream hype, a quieter shift is happening among the platform’s power users.

With the AI arms race heating up, the massive gap between OpenAI and the competition has pretty much vanished, especially for power users like developers and data analysts. Don't get me wrong, OpenAI is still cranking out incredible updates at a crazy pace. The issue is that its rivals have caught up on the basics, and they’re actually starting to beat OpenAI on the exact tools needed for serious work.

• RELATED: Anthropic just filed for IPO: 5 things you need to know

If ChatGPT doesn't sharpen its edge on reliable long-context recall and autonomous multi-agent execution, it risks ceding its most demanding users to Google's Gemini and Anthropic's Claude.

It's no longer a specs race

It’s tempting to look at the latest AI flagship models and assume whoever has the biggest "memory" is winning. Although, even there Gemini Spark and Gemini Intelligence give ChatGPT a run for its money. But memory alone, is where a lot of people get it wrong. The massive context-window gap that OpenAI once dominated has officially closed. Take a look at how the top three stack up today:

• OpenAI’s GPT-5.5: Ships with a 1 million-token context window.
• Google’s Gemini 3.1 Pro: Matches that at roughly 1 million tokens (about 1,048,576 tokens, to be exact) — putting to bed those older 2M rumors from the Gemini 1.5 era.
• Anthropic’s Claude Opus 4.8: Sits comfortably in the exact same heavyweight tier.

The argument is no longer about which chatbot forgets your conversation first. All three of these models can ingest an entire coding repository or a massive 900-page book in a single prompt.

Instead, the battlefield has moved to how reliably a model reasons across that data, and how long it can work on its own without a human babysitting it. And right now, ChatGPT is starting to look merely competitive rather than dominant.

Anthropic's edge is true 'set it and forget it' autonomy

Anthropic’s newly dropped Claude Opus 4.8 is not only smarter, it wants to do your job for you. Alongside the model, Anthropic launched Dynamic Workflows (currently in research preview) for Claude Code. This lets the AI map out a massive project, spin up hundreds of parallel sub-agents to do the heavy lifting, run for hours and double-check its own work before handing it back to you.

Anthropic is backing this up with some serious real-world claims:

• Codebase-scale heavy lifting: Anthropic says Claude Code with Opus 4.8 can execute entire codebase migrations across hundreds of thousands of lines of code, running tests automatically to ensure nothing breaks before asking for a merge.
• 4x fewer mistakes: Opus 4.8 is reportedly four times less likely to let coding flaws slip through compared to its predecessor, Opus 4.7. It’s built to flag its own uncertainty instead of guessing. For power users, that is the difference between an assistant you have to audit line-by-line and one you can actually trust to run unattended.
• Benchmark dominance: On the rigorous Super-Agent benchmark, Opus 4.8 was the only model to complete every single testing case end-to-end—outperforming both previous Claude versions and GPT-5.5.

Google’s edge is seeing, hearing and deep reasoning

Google isn't trying to build a bigger window with Gemini 3.1 Pro; it’s focusing on what the AI can do inside the window it already has.

Gemini 3.1 Pro is built natively for absolute power users. More than just reading text, it simultaneously processes text, images, audio, video and code at a level the competition struggles to match. The 3.1 update specifically targeted software engineering, financial modeling and agent reliability.

If you're a video editor dropping in hours of raw footage, or a financial analyst feeding it a sprawling, chaotic spreadsheet workbook, Gemini’s native multimodal reasoning is incredibly hard to beat. It’s an area where ChatGPT is suddenly forced to play defense.

But don't count ChatGPT out just yet

To be fair, OpenAI isn't exactly asleep at the wheel. They are shipping aggressive updates to combat this exact pressure:

• GPT-5.5 was engineered specifically to "do more with less guidance."
• Codex CLI has evolved into a persistent, autonomous agent featuring a hands-off "Goal Mode."
• GPT-5.5 Instant has dramatically cut down on hallucinations for high-stakes prompts.

The problem for OpenAI isn't that ChatGPT is falling behind or getting worse. But features that used to make ChatGPT stand out as the default option have been matched, and in some autonomous coding metrics, beaten.

A few final thoughts

Honestly, the crown is still up for grabs. If you’re a casual user who uses AI to draft emails, write cover letters or brainstorm dinner recipes, ChatGPT isn't going anywhere and is probably your best option. Just watch out for the syncopation.

But with AI integrated so deeply into our lives, there will soon be a lot more power users who push these models to absolute breaking point. The rubric is changing rapidly. More users are wondering, "Can I hand this AI a massive, multi-hour project and actually trust the final result?"

OpenAI can't coast on speed or minor context upgrades anymore. To keep its most demanding users from jumping ship, ChatGPT's next major leap has to prove it can handle complex, long-horizon tasks on its own and, be honest enough to tell you when it gets stuck.

ChatGPT isn't dethroned just yet, but for the first time in years, it's crown is wobbling.

More from Tom's Guide

• I use AI all day for work — and it's triggering a bizarre sleep phenomenon
• I asked AI how to stop home mold — it told me to ‘Burp’ my house
• I used ChatGPT to adopt a ‘Million Dollar’ mindset — now I'm rewriting my career goals

Whenever anyone asks for the best AI smart glasses to buy, the Ray-Ban Metas are always a default pick. They've been around the block for a while now and have developed into a smart, stylish pair of specs.

However, as we found out at Google’s I/O, things could be about to change with its new Intelligent Eyewear — co-made with Samsung and designed by either Gentle Monster or Warby Parker.

With established design houses like these, you already know the looks are going to be on point. Add Gemini to the mix, and as we’ve already pointed out, it’s not looking great for Zuckerberg.

But is it enough to convince owners of Meta's own specs to upgrade? Let's take a look at what's new and rifle through the rumors to help answer that question.

What will they cost?

Nothing has been confirmed yet on Google's smart glasses pricing — nor do we know the specific launch date (only that they're coming this year). So instead, let's take a look at the Ray-Ban Meta Gen 2s, which are currently going for $379, and offer the full gamut of 3K 60 FPS video capture, nice daylong battery in my use and a timeless design.

Now let's forget Gentle Monster, because glasses coming from a premium design house like this will carry a premium. Warby Parker has that more timeless, accessible feel and vibe, and I'm hoping they can either match or undercut this cost.

This is for the display-free glasses by the way. Google did mention the display glasses will be spoken more about "later this year," but they will come with a hefty price uplift for the Waveguide technology.

Levels to this g(ai)me

Then we turn our attention to what the glasses can do. Because on one side, you've got a respectable Meta AI experience with models that are being upgraded on a semi-consistent basis with new use cases like Live Translation and longer memory contexts to increase usefulness.

But still, while they certainly look impressive in their wayfarer-esque style, you're limited by app compatibility and not being able to talk to other services outside of Spotify and a couple others. Navigation feels a bit unreliable in many locations because of it.

Enter Android XR with Gemini being the main character here, and the difference is night and day. Conversational AI powered by Gemini 3.5 Flash, multi-modal connectivity to all your Google services, and vastly more capabilities than Meta AI at the moment.

Even as a dyed-in-the-wool Apple user, I’m keen to hear more about what the Google “intelligent eyewear” will be like since it’ll be platform agnostic and usable with iOS. Given Apple’s gone a little quiet since Vision Pro’s divisive arrival, I feel like this will be a race to the finish like this fall between Google's eyewear and Apple's own glasses.

My colleague Jason England was able to ask for Google Maps directions, and being able to refer back to Gmail to recall an address or appointment time, or ask Google Calendar to give you a daily summary, could be really useful in just about every scenario — not just the very staid lab demos they had him checking out like a cookbook in another language.

Having generative AI right in front of your eyes won’t sit well with everyone (myself included), but there’s perhaps a spark here that could ignite a broader category in a way we’ve not seen at all.

Outlook

So in my time testing both, one is clear. Google's Intelligent Eyewear has a huge lead in actually useful AI smart glasses over the Ray-Ban Meta. Not to say that the latter aren't useful at all, but it's the ease of use, connectivity to all the Google services you're probably already using and much more that gives them the edge.

Will it be enough to encourage an upgrade from your current glasses and a "jump ship" moment? Possibly, but we'll have to wait for Meta Connect 2026 to find out what new updates may be coming to the Ray-Ban Metas (maybe some OpenClaw support if you're nasty).

More from Tom’s Guide

• I switched my Ray-Ban Metas for L’Atitude 52°N for a month, and while there’s cool features for explorers, it’s no contest
• I put Viture Beast and Xreal One Pro head-to-head for a week: Here's the one pair I'm actually keeping in my bag
• I traveled 5,000 miles with Rokid Glasses — this Meta Ray-Ban Display rival impressed me

We do a lot of stuff on our phones, and having an abundance of storage space can really come in handy — especially if you're finding your old 128GB of free space doesn't cut it anymore. But those storage upgrades can get very expensive if you're not careful, and the worst part is that you can't take them with you next time you buy a new phone.

A 256GB iPhone 17 costs $799, but a 512GB model is $200 more. That's $200 extra that you'll need to pay every time you buy a new iPhone, and all so you can maintain that same level of storage. Luckily, you don't have to pay for extra storage, and there are alternative options available that could save you money in the long run.

Cloud storage

I will caveat this particular point because cloud storage subscriptions are another kind of trap. It's one of the easiest ways to offload files and photos from your phone, but it also locks you into an ongoing subscription that you'll need to maintain for the immediate future. The cheap storage option you start with may start to fill up within a few years, at which point your options are to pay for more storage or end up back at square one.

But investing in a cloud subscription is an easy way to avoid the expensive storage upgrades, at least in the short term. For example, a 200GB iCloud Plus plan is $2.99 a month in the United States, or $35.88 a year.

You don't get quite the same amount of storage as upgrading to a 512GB iPhone 17, but you're also paying considerably less up front. It'd take five and a half years for your total subscription costs to hit the $200 mark, by which point you may be ready to buy a new iPhone anyway.

iCloud isn't the only option for iPhone owners, but it is the most convenient since it's owned by Apple. But if you're willing to forgo that convenience, you may be able to find a better deal with one of the best cloud storage companies instead.

Just remember that you need to maintain your cloud storage subscription indefinitely if you want to retain access. Stop those monthly payments, and you stop being able to upload new files — or in some extreme cases, you may lose access to the stuff you already uploaded.

Trading one recurring cost for another monthly bill is not always the most prudent financial choice you can make. Thankfully, there are other, more cost-effective options available.

It's also worth noting that cloud storage is an excellent tool for backing up your phone automatically, which can be a lifesaver in the event your phone veer gets damaged, stolen, or accidentally erased.

Buy a phone with a microSD card slot

These days, finding a phone with a microSD card slot is easier said than done. There are no premium phones that still offer expandable storage the way some of them used to, but phones with microSD card slots do still exist. One of our favorites is the Moto G 2026, which costs less than $200.

The downside is that these are often some of the cheapest phones you can buy, and have all the sacrifices and compromises that come with paying such a low amount for your phone. In some cases, that even means that you have no higher storage options, because there's only one variant actually on sale.

Then again, consistently picking a phone with a microSD card slot means that you never need to worry about transferring files to a new device ever again. You simply need to pop the memory card out of your phone and slip it into the replacement — a process that takes less than a minute. microSD cards are also portable, easy to upgrade and can be plugged into other devices if needed.

Considering these phones are already on the lower end of the pricing spectrum, you're technically saving even more than you would have done if you bought a flagship device. Just remember that you will need to be happy with the lower performance and weaker cameras that come with cheap phones.

Offload to external storage drives

The most cost-effective way to avoid having to continually buy high-storage phones is to invest in some external storage. Here, there are two routes: a network-attached storage, or a simple portable hard drive.

A network storage system will cost more upfront, but most can be configured so that you can back up your phone automatically to the drive, and then access it via the cloud, wherever you are. NAS drives also have other functions, like being able to store your music and movie collections, so it's more than just a one-trick pony.

The less expensive option is to purchase a portable hard drive or one of the best SSDs, ideally with a USB-C cable. A 1TB portable hard drive can be purchased for under $100 if you go with a lesser-known brand. Even the more well-known names, like Western Digital, Seagate, or Samsung, don't charge a whole lot more for similar levels of storage.

Just about every modern phone or tablet can connect to external storage of some kind, meaning you can plug your drive straight into your phone and offload files without using a desktop machine as an intermediary. Older iPhones with Lightning ports may be a little more awkward to handle, but that's due to slow transfer speeds and the shape of the port itself. But most of your issues will be alleviated with either an adapter or a Lighting-equipped USB cable.

The main downside here is that external storage is typically kept offline. While cloud storage is accessible anywhere, via the internet, offloading photos and files to a hard drive means you need physical access to the drive to reclaim your saved files.

Remote access can be an option with the right equipment and know-how, but it isn't a default feature. Likewise, there are hyper-compact SSDs out there that can snap onto the back of your phone via MagSafe, if you need regular access.

But then again, how often do you urgently need access to older photos and files? I'd wager that it isn't very often, which means external storage isn't quite so bad after all.

Learn to manage your existing storage

It doesn't matter what method you choose to avoid paying a premium for high-storage phones; you should always take the appropriate steps to ensure that the storage you have on your phone is adequate. After all, if you always have enough room on your device, then you don't need to spend money on extra storage — regardless of whether it's an external hard drive or an ongoing cloud subscription.

The best way to do this is to keep your storage requirements as low as possible and not fill it up with superfluous stuff that you don't need.

One part of this is keeping the number of saved photos under control, which can be handled in a number of ways. The main one is to make sure you delete downloaded content that you don't intend to keep, including any memes or files sent over via instant messaging apps. The files themselves may not be very big, but they all add up over time.

It's also possible to reduce the resolution of any photos you take with your camera. While high-resolution photos and videos may sound good, they take up a lot of space and the difference in quality usually isn't that noticeable to the casual observer. For instance, a 12MP photo I took on my Pixel 10 Pro is 1.9MB, while a 50MP shot is 8.6MB.

Shooting in RAW mode also means higher storage requirements versus JPEG. In my case, 12MP and 50MP RAW shots take up 12MB and 35MB, respectively. That single hi-res RAW shot takes up more space than 18 12MP JPEGs.

You also need to remember that mobile apps take up a lot of space, too, as do the various pieces of information they store. It's always worth checking the Apps menu of your storage settings and seeing which pieces of software are using up more space than they should. For instance, my camera currently takes up 2.47GB of space, with 1.09GB vaguely labeled as "User data," while Audible is taking up 3.46GB — 3.26GB of which are various downloaded audiobooks.

Try to give these apps a purge every once in a while, but since you often can't pick and choose which stray files get deleted, make sure that you don't end up erasing something that can't be recovered. Google Photos takes up almost 13GB of space on my phone, but since clearing the storage would also mean deleting precious photos and videos, I leave it well alone. So, if in doubt, let the app take up the space it needs.

More from Tom's Guide

• I tested this $249 phone that has dual screens — and makes me question why we don’t have more of them
• Trump Phone lands with 'dehydrated pee vibes' according to early impressions — here's why the President's freshman effort won't scare Tim Apple
• The best iPhone lenses in 2026

I discovered my Android phone was constantly sharing my location, microphone, camera, and contacts with apps running in the background. Most requests are legitimate — Maps needs location, your camera app needs camera access. But many apps hold permissions I'd completely forgotten granting months ago.

Google's Privacy Dashboard on Pixel phones shows exactly which apps accessed what data and when. If you're using a Samsung device, the feature is called "Permission Manager" and works similarly.

When I launched the dashboard, it was eye-opening — and not a little bit scary. Here's how to access the Privacy Dashboard on Pixel/Google Android, and take back control of your data.

More from Tom's Guide

• This iPhone and Android setting might be quietly leaking your home address
• 3 Android speed killers hiding in plain sight — here's how to fix them
• The most common charging habit is actually killing your phone — here's the fix

A large part of the appeal of Google's Chromebooks has been their lower price, but the new Googlebooks are launching at a premium, which means much higher pricing.

However, in a recent interview with Chrome Unboxed, John Maletis, Google's Vice president and GM of ChromeOS, confirmed that more affordable versions will be available down the line.

“We’ve always been about enabling technology and the ability to be productive and access information regardless of your price point,” Maletis said. “So over time we will come down, but these first devices are super premium.”

Outside of Apple's MacBook Neo, the trend for laptop makers in 2026 has been to ditch their budget laptops in favor of premium options, which command higher prices and avoid the RAM crisis. Unfortunately, that might be working, as The Register reported that Lenovo posted record numbers as it switched to premium PCs and dropped its budget lineup.

Which is why it's good that Google is even promising that more affordable options are on the roadmap.

What are your budget options?

Despite announcing the Googlebook series, Google has promised that Chromebooks aren't going anywhere yet. So the best Chromebooks are still available and an affordable option for those looking for something more wallet friendly.

Besides, Googlebooks are meant to compete with Apple's high-end MacBooks, not the Neo.

While the MacBook Neo is the best budget laptop, there are plenty of other options, including the Lenovo IdeaPad Slim 3x and the Acer Aspire Go 15. Nicely, with the launch of the Neo, we now have options for every operating system from ChromeOS and Windows to macOS.

Maletis declined to provide a potential timeline for the more affordable Googlebook, so we may have to wait. Meanwhile, despite the memory crisis, you do have options.

It might get worse as we get deeper into 2026, though, so if you're in the market for a budget laptop, now might be the best time to pick one up.

More from Tom's Guide

• Biggest Google I/O 2026 announcements — Gemini Spark, Intelligent Eyewear glasses and more
• Samsung Wallet just matched Apple Wallet and Google Wallet on digital passports
• 'Google is force-feeding AI with no way to opt ': DuckDuckGo CEO says installs are surging after Google I/O

Google’s AI Overviews are facing fresh criticism after some users discovered the feature struggling to spell basic words correctly — including the word “Google” itself.

At first glance, the issue almost feels funny. According to Tech Crunch and other reports, AI Overviews returned incorrect letter counts for simple words and generated bizarre responses for searches involving words like “disregard,” “ignore,” and “stop.”

But after using Google’s AI-powered search experience extensively, I think these mistakes highlight something much larger happening with search right now. Because unlike older Google experiments, AI Overviews are no longer optional for many users.

The paradox of AI logic

Ironically, I’ve personally had Google’s AI Overviews help me with spelling and typo corrections before.

Sometimes when I search too quickly on mobile or mistype a word entirely, AI Overviews instantly interprets what I meant and surfaces the correct answer anyway. In those moments, the feature feels genuinely useful.

That’s part of what makes the current criticism so interesting. The same AI system that can intelligently correct a complex typo one minute can completely fail at basic spelling logic the next.

How can a system be so brilliant yet so profoundly dumb? The answer lies in how these models are built. AI doesn't actually read individual letters; it digests language in chunks of words. It can easily understand the conceptual meaning behind a messy typo, but asking it to count individual letters forces it to do math on something it only understands as a "vibe." That deep structural inconsistency is the real issue.

Search is becoming probabilistic

Traditional Google Search was never perfect, but there was a comforting predictability to it. When you searched, Google matched those words against indexed websites. You clicked links and evaluated sources yourself. But AI Overviews has fundamentally changed that.

Now, Google is increasingly attempting to interpret intent, summarize information and generate conversational answers before users ever reach the open web (as indicated in the humorous reel above). That means even tiny inaccuracies can snowball into larger trust problems.

If an AI can confidently misspell a simple word or misunderstand a straightforward query, users naturally start wondering where else those mistakes might appear:

• Medical searches and health diagnoses
• Financial advice and market trends
• Legal information and compliance
• Product recommendations and reviews
• Breaking news and real-time events

And unlike ChatGPT or Claude, many users never explicitly opted into this experience. Google is progressively placing AI-generated answers directly at the top of search results by default. That’s why these errors feel bigger than a meme.

Maybe Google is just moving too fast. To be fair, Google has already acknowledged some of these AI Overview issues and says fixes are underway. But the broader rollout reflects how aggressively the company is pushing toward AI-first search after years of pressure from competing AI platforms.

At Google I/O, the company essentially reframed Search as an AI product rather than just a link engine. The result is a search experience that increasingly behaves less like a library and more like a chatbot.

The takeaway is AI has a confidence problem

The most dangerous AI mistakes usually are not the outrageous ones. It’s the subtle errors delivered with complete confidence.

That’s especially important as more people begin relying on AI-generated summaries instead of clicking through multiple sources themselves. Now that AI Overviews are effectively being pushed onto mainstream users, these small cracks in reliability suddenly matter a lot more.

More from Tom's Guide

• I almost got scammed by AI Overviews — here's how to keep it from happening to you
• Can you trust AI Overviews? Recent studies suggest they may not be as accurate as you think
• Google AI Overview — 3 things I like and 2 I don't

The growing backlash against AI has spilled into nearly every corner of culture — from art and music to education and the future of work. Spend even a few minutes scrolling through social media and you’ll likely come across videos of graduating college students booing pro-AI commencement speeches, highlighting just how skeptical younger generations have become about the technology’s rapid rise.

With AI now dominating conversations across politics, entertainment and Silicon Valley, it’s no surprise the debate has reached one of the world’s most influential religious leaders. In his first encyclical, Pope Leo XIV warned about the dangers of artificial intelligence and stressed the importance of governing the technology responsibly before it reshapes society beyond human control.

The Pope’s letter includes several striking passages about AI, but one quote in particular will instantly stand out to devoted fans of The Lord of the Rings. Seeing a reference that feels straight out of Middle-earth woven into a serious warning about artificial intelligence is not something many people expected from the Vatican.

Pope Leo XIV says AI must be 'disarmed'

In the Pope’s extensive Magnifica Humanitas (Magnificent Humanity) letter, Pope Leo XIV repeatedly emphasizes the importance of “preserving the human person in the age of artificial intelligence.”

“For AI to respect human dignity and truly serve the common good, responsibility must be clearly defined at every stage: from those who design and develop these systems to those who use them and rely on them for concrete decisions,” the Pope wrote in a section titled “Responsibility, transparency and the governance of AI.”

He also touched on many of the biggest debates surrounding AI today, including how rapidly the technology is reshaping everyday life and concentrating power among those who already hold significant economic influence.

At the same time, the Pope stopped short of calling for AI to be abandoned altogether. Instead, he argued that the technology must be carefully controlled before it begins to dominate humanity itself.

“To disarm does not mean rejecting technology, but preventing it from dominating humanity,” he wrote. “It means freeing technology from monopolistic control and opening it to discussion and debate, therefore making it human-friendly and restoring it to the plurality of human cultures and ways of life.”

The Pope added that AI should become more transparent, accessible and thoughtfully governed by the developers and companies building these systems.

One especially powerful line from the letter may sound familiar to longtime fans of The Lord of the Rings and The Hobbit. Referencing words associated with the wizard Gandalf, the Pope wrote:

“It is not our part to master all the tides of the world, but to do what is in us for the succour of those years wherein we are set, uprooting the evil in the fields that we know, so that those who live after may have clean earth to till.”

Using that literary reference, the Pope argued that humanity’s greatest defense against dehumanization will come not from technological dominance, but from “small and steadfast acts of fidelity” rooted in compassion and human connection.

The letter was significant enough to draw praise from Chris Olah, who echoed the Pope’s call for broader conversations around AI governance.

“In conversations we at Anthropic have had with leaders across faith and cultural traditions, we found one shared and deeply held conviction: if this technology is coming, it must go well—for our common home, and for the children to come,” Olah said.

Bottom line

Pope Leo XIV has used his high position to speak out against several concerning matters, such as President Donald Trump’s push for immigration deportation and the war in Iran.

He’s even apologized for the Vatican’s role in legitimizing slavery and not condemning it in the centuries since doing so. To see him also make his stance known on AI and wish for stronger oversight on its continued development is equally fascinating to witness.

More from Tom’s Guide

• 'People will buy intelligence from us on a meter': ChatGPT's CEO, Sam Altman, has critics worried with his AI vision
• Even the people building AI don't know exactly where it's going
• 'Google is force-feeding AI with no way to opt ': DuckDuckGo CEO says installs are surging after Google I/O

If you’re deep into the Google ecosystem like I am, you probably already know about Gems — Google’s custom AI chatbots, similar to OpenAI’s custom GPTs. I’ve been using them constantly because they’re great at locking in a specific tone, persona or workflow so I don’t have to keep retyping giant prompts every single time I want help with something.

However, I recently realized that even with custom Gems, I was still manually opening Gemini, typing requests and waiting for responses. It still felt like I was doing too much of the work. That’s when I started experimenting with Google Workspace Studio, Google’s no-code automation platform, and honestly, it changed the way I use AI during the workday.

Instead of chatting with my Gem directly, I can now plug it into automated workflows that run in the background. That means my AI assistant can summarize emails, analyze files, organize spreadsheets or draft replies without me constantly babysitting the process.

Once I got it set up, it genuinely felt like I had built my own behind-the-scenes, 24/7 assistant.

Whether you want to automate repetitive admin work, create smarter email workflows or build an AI-powered content pipeline, here’s exactly how I started using custom Gems inside Google Workspace Studio.

What I needed before getting started

Before I could connect everything, there were a couple of requirements I had to figure out first.

• The Google Drive rule: I learned pretty quickly that my custom Gem had to be grounded using Google Drive files like Docs, Sheets or PDFs. If the Gem relied on outside extensions like YouTube or Google Maps, it wouldn’t appear inside Workspace Studio.
• Admin permissions: Workspace Studio is more of an enterprise-level tool, so, if you're using Gems on your work computer, Gemini actions need to be enabled by your organization’s Google Workspace admin before everything can work properly. If you use Workspace Studio on your own, than you are the admin and don't have to worry about this.

If you want to see if your current account supports Workspace Studio, log into your computer and go to: studio.workspace.google.com. If it lets you in, you can can start using Gemini to build automations (like telling AI to auto-summarize unread emails, or automatically saving specific Gmail attachments to a Drive folder).

If it blocks you or shows an error, it means you are either logged into a free personal Google account, or you are on a corporate Workspace account where your company's IT administrator hasn't enabled Gemini features yet.

Once you're in, you're ready to go.

How I added a custom Gem to an automated Flow

• Sign in and click: "Create New Flow" to build a workflow from scratch.
• I used a template. Every Flow needs something that kicks off the automation. For this, I just used the templates Google already offers. They are pretty good, but if you prefer to set your own, simply type your rule in the chat.
• Use cases. I used mine for Gmail, but you could set a flow for Google Drive uploads, form submissions, spreadsheet updates, calendar events or really anything when it comes to "if this happens..." part of the workflow. (i.e. if I get a form submission, add it to Google Sheets).
• I added the 'Ask a Gem' action. Next, I clicked the blue + Add step button underneath the trigger. Under the AI actions section, I selected Ask a Gem. This is where the magic happens because it lets your custom Gem process information automatically in the background. Then, connect your custom Gem.

Once you have everything set, you're going to want to test everything. Luckily, Google lets you click "Test run" to make sure the Gem handled the information correctly. If it doesn't, you can ask Gemini for help. You can ask within the space, or if you feel more comfortable, take a screenshot and use the app. If you're like me, sometimes it helps to seperate into a new app just to avoid confusion.

When my Flow worked, I switched the Flow to "Turn on" and let it run automatically.

For me, the coolest part was seeing how the Gem responds because that response then becomes usable data for the rest of the workflow. In other words, instead of of stopping after the AI does the job, I could then chain additional actions afterward.

For example, I used my Gem’s output to:

• Gmail: Automatically create draft replies I could quickly review and send
• Google Docs: Generate organized summaries and reports inside templates
• Google Sheets: Route parsed data and AI analysis directly into tracking spreadsheets

Honestly, this is where it felt both too easy and too good to be true. Gemini became the admin assistant I've always needed. And even better, it is so easy to set up. Once you log into the Workspace Studio, the interface is clean and simple enough that setting up a Google Gem almost feels intuitive.

And Gemini is just a click away, so you can always ask it questions or tell it what you're trying to do. If you turn on Gemini Live, you can even share your screen.

Bottom line: I should have done this sooner

Gemini 3.5 Flash makes the entire user experience feels effortlessly fast and seamless. That's why it's a great time to explore what Gemini can do outside of the chat window. And trust me, if you can prompt Gemini, you can make the most of AI in Google Workspace Studio for autonomous productivity.

Once you get a taste of what AI can help you work with, I have no doubt that you'll increasingly start teaching the AI systems how to handle recurring tasks for you automatically.

And after trying it myself, I don’t think most people realize how close we already are to having fully personalized AI workflows — not to replace us, but assist us so we can focus on the more important stuff.

More from Tom’s Guide

• I used the ChatGPT ‘butter’ prompt for deep research — and my results got way better
• ChatGPT might be quietly rewarding people who know how to think clearly — these prompts can help
• I tested ChatGPT vs Claude vs Gemini to build a resume with Canva — and there's a clear winner

Environmental activist Erin Brockovich, best known for exposing a major water contamination case against Pacific Gas and Electric Company that later inspired the Oscar-winning film "Erin Brockovich," is turning her attention toward AI. Brockovich is specifically sounding the alarm about massive data centers powering tools like OpenAI's ChatGPT, Google's Gemini and other generative AI systems.

According to a recent report from Newsweek, Brockovich is asking Americans to help crowdsource information about proposed and existing AI data centers across the country through a new public map initiative.

At first glance, it might sound niche. But the implications are much bigger than they appear.

What is Erin Brockovich doing exactly?

Brockovich is shifting the conversation of how AI is framed. Rather than as software such as chatbots, image generators and productivity tools, she's emphasizing the enormous industrial infrastructure required to power these tools.

From increasing electricity demands and rising water usage to grid strain and giant server farms, Brockovich is highlighting the fact that the hidden cost of AI should not be ignored.

Unfortunately, most people don’t think about what happens behind the scenes when they ask ChatGPT a question or generate an AI image. But every AI prompt requires real-world computing power — and a staggering amount of it.

Companies including Microsoft, Amazon, Google and Meta are now racing to build more data centers to support increasingly powerful AI systems. But that boom has triggered growing concerns around land use, water consumption for cooling systems and energy demand.

In some parts of the country, communities have already begun pushing back against proposed projects over fears that AI infrastructure could dramatically reshape neighborhoods and strain local resources. But that’s what makes Brockovich’s involvement so significant. She helped turn environmental contamination into a mainstream public issue decades ago. Now, she appears to be trying to do something similar with AI infrastructure.

AI is starting to feel physical

The dispute about where to build data centers takes AI from softerware to something much more tangible, essentially, a massive industrial system with real-world consequences.

Brockovich is helping people start connecting their rising utility bills, local development projects and water concerns to AI expansion. The hope is that this awareness will help the public conversation around AI change.

Interestingly enough, the conversation, especially towards the opposition to large-scale data center expansion is increasingly bipartisan. Environmental groups are raising sustainability concerns and local residents are worried about noise and infrastructure strain. Some conservatives have also pushed back against large tech developments reshaping smaller communities.

The debate is turning to questions about who benefits from AI, who pays the cost and whether communities should have more say in how AI infrastructure expands. Yet, AI companies continue investing billions into new computing facilities. In other words, it doesn't look like AI is slowing down, so we can expect these conversations to become much more common.

Final thoughts

The AI race is everywhere. It's moved away from Silicon Valley to real concerns happening in neighborhoods and local communities across America. Now with Erin Brockovich now entering the conversation, scrutiny around AI infrastructure may be entering an entirely new phase.

More from Tom’s Guide

• I used the ChatGPT ‘butter’ prompt for deep research — and my results got way better
• I tested ChatGPT vs Claude vs Gemini to build a resume with Canva — and there's a clear winner
• ChatGPT might be quietly rewarding people who know how to think clearly — these prompts can help

I don't mean to brag, but aside from a little photosensitivity, my eyesight is damn near perfect. In the 30-plus years I've been alive, I've often been grateful for the fact that I don't need any sort of corrective eyewear to function. Unfortunately, in a world increasingly interested in the best smart glasses, that poses a problem.

I imagine that if I did wear glasses, I might be more enthusiastic about smart glasses as a whole. It doesn't matter whether we're talking about specs with built-in displays, like Meta Ray-Ban Display, or something that relies on voice and audio like Intelligent Eyewear from Samsung and Google. But the fact that I generally don't wear accessories on my face means the entire prospect of smarter eyewear does not appeal to me in the slightest.

I say this following the reveal of Samsung and Google's Intelligent Eyewear at Google I/O earlier this week. The glasses may not have displays or gesture controls, but the addition of Gemini and its connection to Google apps and services has wowed some of my colleagues here at Tom's Guide. And yet, my lack of eyewear experience means I can't really see the appeal.

What do smart glasses do?

Different smart glasses offer different features, depending on who made them and on the hardware available. Typically, these frames come with an integrated camera, microphone, personal speakers, and a way to interact with an AI assistant. For Meta Ray-Bans, that assistant is Meta AI, while the new Intelligent Eyewear from Samsung and Google is linked in with Google's Gemini.

While the basic hardware allows these glasses to be used as cameras or headphones, the extra capabilities are wholly dependent on the AI you have access to. Like smartwatches, they are essentially a way of interacting with another device, usually your phone, using voice prompts and commands.

Google revealed that Intelligent Eyewear will have access to a whole host of Google services, including Google Maps, which should give it a slight edge over platforms that don't have such a robust software suite to fall back on.

More advanced smart glasses have pushed the envelope a little further, with new hardware and ways of interacting with the glasses themselves. The Meta Ray-Ban Displays include a wristband that enables gesture control, as well as a small screen in your line of sight that adds visual information and apps to the mix.

Fully-fledged AR glasses are also available, with more robust eye-level screens. A good example is the Viture Beast, which is our current pick for the best smart glasses. However, these devices are a lot larger and more obvious, on account of the AR display needing more hardware to operate.

AR glasses certainly feel more appealing to me, as a non-glasses wearer, especially if they can overlay TV or monitor-style screens directly in my line of sight. But AI-only glasses are a much harder sell.

I already don't like using wearables

So I don't wear glasses to correct my vision, that doesn't mean I can't wear smart glasses anyway, right? Sadly, there are two main problems to contend with here. The first is that I don't have much interest in wearing glasses as a fashion statement, as was the trend 10-15 years ago when hipster culture seemed to be everywhere.

Wearing frames with blank lenses, or no lenses at all, is going to make me feel almost as ridiculous as if I had a curly moustache tattooed onto my index finger. Not to mention the fact that they're going to need cleaning regularly. I have to clean my sunglasses at least once a day, and I can't imagine having to voluntarily do that more frequently. I feel your pain, glasses wearers, I really do.

I'm also very aware of the fact that I don't really like wearables either. I've used a few smartwatches and fitness trackers over the years, and without fail, I will eventually stop wearing them because of the inconvenience. I know there are benefits to having a smartwatch, but the benefits don't outweigh the hassle of wearing and charging them.

The idea of doing this for a pair of glasses, and wearing them in a position that's almost impossible to ignore, is an instant turn-off for me. Especially when I remind myself that I don't actually like talking to AI assistants or chatbots, and prefer to do all my phone activities manually.

I can see the appeal for some, but not me

I can see why some people may find smart glasses an appealing prospect. If you're already wearing glasses on a regular basis, then the prospect of throwing some smarts into the mix might be quite appealing. If you're going to be wearing glasses to see, then you might as well let them multitask and add a little more convenience to your life.

Likewise, if you wanted to be able to take advantage of Gemini Live or other AI modes that benefit from an active camera feed, then face-mounted cameras make that whole process a lot easier.

But that doesn't appeal to me. I'm not the biggest fan of using AI in the first place, and without AI, those glasses are mostly pointless — especially if you don't need corrective lenses.

Plus, I've lost count of the number of sunglasses I've lost in various dumb and unlikely circumstances. I once lost a pair of sunglasses cycling over an intersection too fast, only realizing it after they'd been run over and destroyed. I absolutely do not trust myself with expensive sunglasses, let alone a hi-tech pair with electronics and a battery inside.

So Intelligent Eyewear might have a bunch of cool and interesting features, but they are definitely not going to earn a place on my wish list.

More from Tom's Guide

• How to tell if someone is filming you wearing smart glasses — the signs to watch out for
• I just tested Xreal’s Android XR glasses, and they are an early glimpse of my dream future smart glasses
• I wore the Meta Ray-Bans to a Yankees game — here's where they excelled, and where they need work

As a New Yorker who was born and raised in Queens, I quickly learned what separates a mediocre slice of pizza from a truly amazing one.

Over the years, I’ve taken the train and bus just to fulfill my appetite by munching on a Grade A slice from the likes of Cuts & Slices, Macoletta and New Park Pizza. My never-ending journey for the best pizza in the city, best known for its slim slices, eventually lit a lightbulb over my head and pushed me right back into the world of AI.

Ever since Gemini got attached to Google Maps, it’s become a handy tool for those times when I need hands-free directions via chatting while I’m driving, a heads-up about the traffic conditions for my next route and an identifier that gives me verified info on any special places I come across. It’s also great on the recommendation front, which is why it’s become my best buddy when I need to know where to eat next.

When I used the “Ask Maps” chat feature with Gemini in Google Maps and requested a list of the best pizza places in NYC, it mentioned a popular tourist hotspot. My expectations for those sorts of places tend to be low, but to my surprise, it was a lot better than expected.

Making my way to Joe’s Pizza

Once I verbally demanded that Gemini in Google Maps put me on the path to cheesy, tomato-fueled heaven, it brought up a bunch of highly-rated pizza spots from every corner of the city.

It even split them up into different classifications, such as “Modern Champions" (best pizza in the city AND the world!), “Historical Icons” (legendary pizza spots that are cornerstones of NYC) and “Classic Favorites” (go-to spots every NY'er should go to) to make my next pizza destination a bit easier to discern. Thankfully, it didn't bring up any pizza chains like Pizza Hut (which is a plus when you remember AI hasn't had the most positive effect on the popular restaurant brand as of late).

The main pizza spot that caught my attention was Joe’s Pizza, which Gemini offered a very enticing description of stating, "For the ultimate classic NYC slice, Joe’s Pizza is the gold standard." It also stated that it's known for:

• Fame: Famous for its on-screen cameos (including Spider-Man 2) and photos of celebrity fans lining the walls.
• The Slice: Known for thin, foldable slices with just the right amount of grease and a perfect cheese-to-sauce ratio.
• Convenience: Open until 3 AM or later, making it a legendary late-night destination.

It must have been fate that Joe’s Pizza ended up catching my eye because it turns out that it’s a block away from the office here in Times Square. So on a rainy and unusually chilly Thursday, I stepped away from my desk and headed downstairs to see if all the hype about Joe’s Pizza was justified. After waiting 30 minutes in a long line full of locals and out-of-towners, I left with two slices and high expectations.

Two slices of greatness

As someone who can be overly judgmental about pizza (Sbarro’s is not, ladies and gentlemen!), I came away from my mid-afternoon pizza journey with high marks for Joe’s Pizza. Gemini in Google Maps’ lofty overview of what makes the actual location such a cool destination and why its pizza tastes so good were on the money — those two slices I ate made the long wait in the rain worth all the trouble. Once again, AI performed admirably when the time came to lead me to another delicious location that I'll be sure to mention to my fellow pizza addicts.

Final thoughts

Having a quick chat with Gemini in Google Maps just so I can stuff my face is now one of my favorite pastimes. Asking it to point me in the right direction of 4-5 star pizza restaurants is one of the best ways to use Google’s chatbot on the Google Maps app.

After finding out just how good Joe’s Pizza is based on Gemini’s lofty promises, I’m more keen on trusting its other pizza recommendations. It’s safe to say that I’ll be making a few stops to Una Pizza Napoletana, John’s of Bleecker Street and Di Fara Pizza real soon.

If Gemini in Google Maps says it's good, who am I to question it?

More from Tom’s Guide

• Pizza Hut franchisee says AI delivery system cost them millions and 'pummeled consumer satisfaction' — now there’s a $100 million lawsuit
• I asked ChatGPT to apply Asha Sharma’s leadership mindset to my life and it made me completely rethink my priorities
• I asked ChatGPT for a copycat Chick-fil-A recipe — but it left me wishing I'd ordered takeout

Update: Google is aware and working on a fix

At Google I/O 2026, the tech giant revealed the biggest update to Google Search in over 25 years that pushes AI into everything. It's a move that appears poised to transform how the internet has worked for decades.

As my colleague Amanda Caswell wrote, "the internet increasingly feels like something else entirely: one giant conversation happening inside a single AI response box.

What was unexpected is that it looks like Google's AI is also breaking Search itself. Google's controversial AI Overviews appear to be breaking when you use known AI commands like "disregard", "dismiss", and "ignore."

We first started noticing social media posts about the issue where people found searching "disregard" doesn't show the definition of the word. Instead, it says things like "No problem at all. If you have any other questions, let me know."

This is what AI bots tend to do when you're telling it do something instead.

As a reminder, part of the Google Search overhaul involves turning Search into an Gemini-powered AI chatbot that can build you an interactive website, set your calendar, and more.

It's a mixed bag

Several people on Tom's Guide and myself tried searching the known AI prompt words and it's an inconsistent experience.

For myself, on desktop, I have yet to get the broken AI overview. However, when I tried it on my phone, disregard broke the feature but ignore and dismiss didn't.

Tom's Guide U.S. editor in chief Mike Prospero found the words borked the overview in AI Mode including the new words "cancel" and "stop." AI Mode told him "No problem. I've stopped the current action."

My colleage Tony Polanco noted a problem on his work account but his personal one didn't have any issues.

Obviously, not everyone is just searching AI prompts in Google, but plenty of people utilize the built-in dictionary feature as a habit. On the whole, you probably won't run into this issue.

However, it does show the flaws in Google turning over Search functions to AI, especially after the original AI Overview controversies when it was introduced in 2025. Just a month ago, a study found that AI overviews are wrong 1 in 10 times, and could be even worse.

Tom's Guide has reached out to Google for clarity on the issue and I was told that Google knows.

“We’re aware that AI Overviews are misinterpreting some action-related queries, and we’re working on a fix, which will roll out soon," a spokesperson said.

They clarified that it appears to be an issue with how AI Overview interprets certain queries, and were keen to say that it's not related to the new Search tools announced at I/O this week.

More from Tom's Guide

• Biggest Google I/O 2026 announcements — Gemini Spark, Intelligent Eyewear glasses and more
• I tested Google’s “Intelligent Eyewear,” and found the smart glasses that will defeat Ray-Ban Meta
• Google’s new $100 AI Ultra plan just changed the AI race — here's what you get

The AI race just took another dramatic turn this week.

President Donald Trump reportedly backed away from signing a major executive order that would have created new voluntary guardrails for advanced AI systems — a move that signals the U.S. may now prioritize AI acceleration over regulation.

According to reports from Reuters and The New York Times, the order would have encouraged companies like OpenAI, Google DeepMind and Meta AI to share powerful AI models with the federal government before public release so agencies could evaluate potential national security and cybersecurity risks.

But Trump reportedly pulled the plug after concerns that the order could slow down American AI innovation while China rapidly advances its own AI ecosystem. For that reason, this may be one of the strongest signals yet about where AI in America is heading next.

The AI safety debate just shifted again

For the past two years, much of the AI conversation in Washington has revolved around safety. Conversations about whether AI companies should be forced to test powerful models like Anthropic's Mythos before launch. And, how the advancement of AI could threaten jobs, cybersecurity and even elections. Also, how to prevent deepfakes and misinformation.

The proposed executive order appears to have been designed as a middle ground. Reports suggest it would have created voluntary cooperation between AI companies and the government rather than hard regulation.

Trump reportedly worried the order could act as a “blocker” for U.S. AI companies competing against China — reflecting a growing belief among some tech leaders that America’s biggest AI threat isn’t unsafe AI, but losing the AI race altogether.

What this means for AI tools like ChatGPT and Gemini

For consumers, this likely means AI companies may continue moving extremely fast. Users could see more experimental AI features rolling out quicker, faster updates, more autonomous “agentic” AI systems and less federal oversight over how these systems launch.

Silicon Valley may quietly welcome this because AI leaders increasingly believe regulation itself could become a competitive disadvantage. If U.S. companies face stricter oversight while rivals overseas move faster, some fear America could lose its lead in foundational AI models.

That’s especially important as AI increasingly becomes tied to national security, military systems, economic growth, scientific research and infrastructure.

The concern from critics is that innovation may now outpace safeguards. According to the Center of AI Safety, the acceleration of artificial intelligence is much faster than the safety research, which could lead to catastrophes.

The result is a growing divide between “AI safety” advocates and “AI accelerationists” who believe speed matters more than caution. And this latest decision by the Administration suggests the accelerationists may currently have the upper hand.

The bigger issue nobody can fully answer yet

The truth is nobody really knows what the correct pace of AI regulation should be. Move too slowly, and governments risk losing control over increasingly powerful systems. Yet, move too aggressively, and you may slow down innovation in one of the most transformative technologies in decades.

That tension is what now sits at the center of nearly every major AI debate happening in Washington and Silicon Valley. And after this latest move, it increasingly looks like the U.S. government is leaning toward one core idea of win first and regulate later.

Bottom line

Trump backing away from this AI executive order may end up being far more significant than it initially sounds. This move signals that the future of AI policy in America may increasingly revolve around competition and speed rather than precaution and safety.

For everyday users, that likely means AI tools will continue evolving rapidly — with more powerful features arriving faster than ever. Whether the safeguards keep up is the question nobody can answer yet.

More from Tom’s Guide

• 5 signs you’re still using ChatGPT like a beginner — and how to fix them
• I used NotebookLM to make a 'Family Wiki'— and now everything I need to run the household is a click away
• Google’s new $100 AI Ultra plan just changed the AI race — here's what you get

Google Search used to feel like opening a door to the internet. There was a time when we could type something in the Search bar, hit enter and disappear down a rabbit hole of websites, forums and Reddit threads. We'd see random YouTube videos and blogs made by people who we truly wanted to know. The pure serendipity of the experience was fun and interesting.

My sister actually met her husband that way. She Googled "Mr.Show" 20 years ago and stumbled upon her now husband's website about the sketch comedy show starring Bob Odenkirk and David Cross. I still love hearing them tell that story.

Now, after Google’s latest AI search updates, the internet increasingly feels like something else entirely: one giant conversation happening inside a single AI response box.

Google's AI-powered vision for Search

At Google I/O 2026, Google doubled down on its AI-powered vision for Search with expanded AI Mode features, conversational follow-ups, synthesized answers and tools designed to keep users interacting with Google’s AI instead of bouncing between websites. The experience looks fairly smooth and easier to digest than the traditional wall of links.

Instead of opening 15 tabs to compare information yourself, Google increasingly does the comparison for you. Now, rather than hunting for the best answer, Google just generates one. In other words, the days of "surfing the web" are over.

That's a shift I know I'm not ready for. But, I've been Googling long enough to know that Google Search has never been ideal. For years, people complained about results crowded with SEO spam, low-quality affiliate content, endless recipe backstories and websites clearly written more for algorithms than humans. And while AI search solves a lot of that, it's forced discovery and curiosity out of the picture almost entirely.

A decade ago we weren’t consuming content at the rapid pace we do now, we were exploring spaces created by actual humans. Now AI increasingly compresses all of those voices into one unified response. The more AI summarizes the web, the more the web itself risks losing its personality. That’s where the “group chat” feeling starts creeping in.

AI controls the conversation

In a group chat, information moves fast. As a mom in a group chat with 14 other moms, I know just how hard it is to keep up. Everyone talks over each other even if they don't mean to, context gets compressed, the loudest voices dominate and most people don’t leave the conversation to independently explore sources. That’s increasingly what AI-powered search feels like.

The internet used to feel expansive — now, it feels summarized. The irony, of course, is that AI-generated answers still rely heavily on websites. Publishers, creators, journalists, bloggers, reviewers and independent writers are still doing the work of reporting, testing, researching and publishing original information. AI simply reorganizes it into a cleaner conversational layer.

Unfortunately, users may never actually visit the sources anymore. If Google becomes the destination instead of the gateway, the entire relationship between websites and readers changes.

And honestly, users may not even notice it happening because the new experience feels so much easier. And that's the tradeoff at the center of this AI era. Convenience, speed and summarizations trump exploration, discovery and individual voices.

The walled-garden paradox

Statistically speaking, you probably didn't type "tomsguide.com" into a browser bar to find this page. You likely found it because an algorithm — quite possibly Google Discover or a personalized Google feed — served it up to you on your phone based on your reading habits.

This highlights the bizarre paradox of the internet today. To read a critique about Google enclosing the open web into a walled garden, you have to pass through the gates of that very same walled garden.

We are relying on the gatekeeper to show us the article complaining about the gatekeeper. Right now, that system still works in a fragile sort of harmony: Google’s feed sends you to our website, you read our writers and we get the support needed to keep reporting.

Final thoughts

Maybe this is inevitable. Maybe AI assistants replacing traditional browsing is simply the next phase of the internet, the same way streaming replaced video stores and TikTok replaced cable TV for an entire generation.

But something undeniably feels different now. The internet used to feel like a giant world waiting to be explored. Now I just feel like I'm chatting with AI.

And websites like this one? Well, they’re at risk of becoming background characters in a chat they helped create. So if you enjoyed this article, I hope you come back to the site. Bookmark the page and comment. Remind us journalists that we still matter in a sea of AI.

More from Tom’s Guide

• 5 signs you’re still using ChatGPT like a beginner — and how to fix them
• I used NotebookLM to make a 'Family Wiki'— and now everything I need to run the household is a click away
• Google’s new $100 AI Ultra plan just changed the AI race — here's what you get

The next generation of Android Auto was announced at the Android Show, and I got to properly test it at Google I/O. In short: Apple CarPlay feels washed in comparison — let me explain.

It takes the car from simply being something you go from Point A to Point B in, and adds an additional layer of making it a central hub for navigation, productivity on-the-go and entertainment. And with Gemini thrown in, Cars with Google built-in just took an evolutionary agentic AI step too.

So I took a (simulated) ride in the Kia EV9 with Android Auto and the full Google built-in Volvo EX60 to see what the upgrades bring to the table.

On-the-go

It all starts with the biggest update to Google Maps in over a decade — Immersive Navigation provides a nice 3D view with a huge increase in details like lanes, traffic lights and stop signs. Being able to see everything with greater clarity like this is a huge step forward for getting around.

Then there’s the multi-widget layout of the new Material 3 Expressive design, which creates a really nice, refined UI for the car and multiple app widgets for quick glance interactions.

Top it all off with Gemini and even agentic actions like ordering a meal on Doordash for when you get home (and even adjusting Google Home controls), and it simplifies the idea of multitasking while you drive with the power of your voice — all in a way that Apple CarPlay has never been able to do.

Stuck in traffic

But if there’s one thing I’ve learnt while out here at Google’s campus, it’s that you spend most of your time in California stuck in traffic not really doing much at all. The standstills are painful, and if you’re in an EV, you could be spending up to 30 minutes sat in your car waiting for it to charge.

That’s where the redesigned audio experience in apps like Spotify come in, and most importantly, support for full HD video up to 60 FPS on the center console. Throw in Dolby Atmos support and the surround sound experience was beefy and all-encompassing. But of course, this is only when you throw the car in park — once you start driving, that video vanishes in favor of an audio-only overlay.

Given the explosion in video podcasting on YouTube, I can see this being a hugely-adored feature for the moments you’re parked in traffic for the full visual experience, and switching on the fly to audio when you start crawling forward.

Your own personal car butler

But things get really interesting when you move over to Cars with Google built-in. For those uninitiated, most cars support Android Auto through connecting your phone to the car and project the content of it on-screen.

Cars with Google built-in are natively-built on an Android Automotive operating system — operating entirely as standalone devices that don’t require a phone to be connected to it. They’ve been around since 2020, and now we’re getting the biggest updates to these models since the launch.

The big thing here is Gemini is able to tap into your car’s controls and onboard sensors. No, that doesn’t mean it’ll drive the car for you, but what it actually means is Google Maps can be far more accurate and give real-time advice by using the car’s sensors to identify which lane is needed for your exit.

On top of that, you can ask Gemini to adjust pretty much all the settings in your car, like making the sun roof transparent, changing the climate control, and altering the mood lighting of the cabin. Rather than getting lost in the many panels of settings, having an AI assistant that can actually take action means less time looking at the screen and more time watching the road.

And even better, it’s got some conversational powers too to help you make sense of the world around you. For example in the demo I took part in, we had a first-person video of a drive through San Francisco, where we asked Gemini to identify the big tall building ahead. Google then went straight to the cameras, identified it and gave me an answer in seconds.

Outlook

So far, I’ve always felt that smartphone-enabled in-car experiences have always been a mask over the frankly terrible software that carmakers subject you to. They’ve always been handy, but only so far and you had to be ripped out of the experience elsewhere.

And while that’s still the case with Android Auto from your phone, Cars with Google built-in finally feel like the revolution that’s needed in this space to make car software actually fun to use — and the secret to it is Gemini intelligence.

Taking the confusion out of your car settings is a huge step forward alongside Immersive Navigation, a new cleaner Material 3 Expressive design and the ability to play full HD video for making your motor a central hub wherever you are.

It’s yet another massive step forward for Google that’s leaving Apple’s CarPlay in a serious need of catching up. All eyes on WWDC 2026!

More from Tom's Guide

• I hate that I love Google’s new AI-powered search bar — it makes the internet easier to digest, but my career is cooked now
• I tested Google’s “Intelligent Eyewear,” and found the smart glasses that will defeat Ray-Ban Meta
• I just tested Xreal’s Android XR glasses, and they are an early glimpse of my dream future smart glasses

Google just rolled out Gemini 3.5 Flash with upgrades that you're going to want to put to the test in your everyday workflow. So much more than a typical chatbot, Gemini 3.5 Flash isn't just fast, it's a true digital assistant that can actively help manage parts of your life — specifically, the boring, time-consuming ones.

Google's latest default model can now work across your inbox, files, photos, travel plans and even coding projects. If you haven't tried this model yet, I have seven practical reasons why it's worth your time.

1. Turn Gemini into an AI executive assistant for your inbox

Most of us waste an incredible amount of time sorting through emails that don’t actually matter. I'll be embarrassingly honest: Before I used Gemini 3.5 Flash to organize my inbox, I had over 10,000 unread personal emails. Because Gemini integrates with Google Workspace, you can ask it to summarize important emails, identify action items and help prioritize your day, which will instantly help you breeze through that unread stack.

Try a prompt like: “Check my Gmail from the last 24 hours and give me the 5 most important things I need to know, along with any action items.”

2. Take a photo of your fridge and let Gemini plan dinner

One of the things nobody tells you when you become a parent is that your kids will be hungry all the time. Just this week, my middle schooler invited his three best friends over to our house after school. All the snacks and drinks I had purchased three days earlier were gone, completely gobbled up by the boys. On days when I feel like the fridge and pantry don't have any ingredients for a real meal, Gemini 3.5 Flash instantly comes to the rescue.

Simply upload a photo of your fridge, pantry or kitchen shelves and ask Gemini to generate recipes based only on the ingredients it sees. You can also mention that you have picky eaters or any dietary requirements.

After uploading a photo or two, try the prompt: “Analyze this photo and give me three quick healthy dinner ideas using only these ingredients.”

3. Upload giant PDFs and ask questions instead of reading everything

AI has long been able to summarize documents, websites and emails. But now that Gemini 3.5 Flash is smarter and faster, you can upload several different PDFs at a time including presentations. Instead of reading through lease agreements, insurance documents, financial reports or even school paperwork, you can upload the file directly into Gemini and ask targeted questions.

For example: “What happens if I break this lease early?”

Or:

“Summarize the biggest financial risks outlined in section 4.”

After that, don't forget to save the document in your own Family "Wiki" to keep everything organized.

4. Build simple apps without knowing how to code

Google is also leaning heavily into “vibe coding,” where users build projects using plain English instead of programming knowledge. To build just about anything, all you need to do is prompt. Using Gemini and Canvas, users can create small interactive web apps simply by describing what they want.

Gemini can then generate the code, create the interface and allow you to test the app directly inside the workspace. This is one of the clearest signs that AI is lowering the barrier to software creation for non-programmers.

For example a vibe coding prompt might be: “Build a simple water tracking app with a visual progress bar.”

5. Use Gemini as a smart travel planner

Looking to get away this Memorial Day weekend or planning a longer trip this summer? Gemini 3.5 Flash can now combine live Google Search information with itinerary planning to streamline the process.

The AI can help organize restaurant recommendations, timing, local events, attractions, travel flow and just about anything to essentially function like an AI travel agent.

A prompt for something like this might look like: “I’m going to Austin for a 3-day weekend. I love live music and tacos but hate long lines. Build me an itinerary and check which venues are open Friday night.”

6. Turn meetings and lectures into summaries and follow-up emails

It seems like everywhere I turn these days there is a transcription tool. And while most of them give a summary, I actually prefer uploading the notes and audio to Gemini. Because the AI now supports audio uploads, it opens up useful productivity features. From voice memos and meeting recordings to brainstorming sessions and lectures, Gemini can extract action items and, my personal favorite, generate follow-up emails. This one is a massive time-saver for professionals and students.

A prompt to try this could be: “Summarize this meeting and draft a follow-up email with next steps.”

7. Get real-time explainers for complicated news stories

Even though I read the news all day and cover a lot of it, keeping up with the latest stories isn't easy. Not to mention, understanding it all through different perspectives. That's why I think one of the strongest uses for AI remains breaking down complex topics into plain English.

Because Gemini connects to live Google Search data, it can explain current events using up-to-date information.

I use the prompt: “Explain what's happening with AI regulation right now like I'm a smart 12-year-old. Include the biggest developments from this week.”

Leaning into a different kind of AI

There's a big shift happening with AI right now. AI is slowly evolving from something that simply answers prompts into something that actively manages workflows, understands context and helps coordinate everyday tasks.

I personally still like to use Gemini 3.5 Flash as a chatbot because I don't need to give it the keys to my entire workspace as an AI agent. Sure, it's capable of that, but using it to understand my inbox, calendar, pull up photos and help me break down documents means I stay in control while also letting it behave more like an operating layer sitting across my digital life.

Give these seven capabilities a try and see what a difference this new model makes with classic workflow problems. Let me know in the comments what you think.

More from Tom’s Guide

• I sold band merch for some of the biggest names in music — here’s why AI can’t replace real artists
• I used NotebookLM to make a 'Family Wiki'— and now everything I need to run the household is a click away
• I use the ChatGPT ‘fart’ prompt whenever I’m stuck — and it somehow works every time

So Google search just got the biggest upgrade in nearly three decades at Google I/O — killing off your standard page of links and replacing it with a redesigned search box giving way to AI agents and a generative UI.

I got to take a closer look, and I must admit it’s a love/hate relationship.

On the one hand, this is a generational leap for search that will surely make the ever-crowded internet much more personalized and curated to everyone using it. But on the other, it’s another potential death blow to tech journalism — or even journalism at large. I’m so conflicted, and therefore need to rant in written form to figure out where my head is on it. So please, join me, won't you?

Making search make sense

The two live demos I saw with Google were keeping up-to-date on indie concerts in my local area and planning a weekend. The former was done on a phone with the latter on a laptop, and the results were seriously impressive.

It’s easy with a rudimentary page of links to fall down clickbait rabbit holes and not actually get an answer to the original question you had, and to see an AI agent work to not only ask you clarifying questions to specify the search but also keep abreast of the information proactively makes this feel like a generational leap. Shout-out to even jumping on Reddit's rumored indie concerts sub too, so you never miss a beat!

Then there’s the generative UI element of the weekend planner, which takes context and memory from what Google already knows about you to build a schedule that knows your tastes and availability. Plus you can share these as pages that others can see too — it’s a really nice, curated way to scythe your way through what is becoming an increasingly noisy internet.

We’re so cooked

But here’s where my problems lie, and I must take you behind-the-scenes a little of how a business like Tom’s Guide works. Like many tech publications, we rely on eyeballs on our website — ads give us a decent chunk of revenue, while the rest is made up of sponsorships and affiliate revenue (when you buy a deal we recommend, we make a little on the side).

Talking in the media lounge, I noticed a definite anxiety amongst my friends from other publications. AI overviews have already shown a sharp decline in traffic referrals to many sites, and by eliminating links for an AI-powered search agent, it’s looking very likely that this downward trend will continue.

In fact, after asking Google about what their expectations are for referral traffic, the answer was essentially to repeat that “more people are searching than ever before,” which is great — but it doesn’t really answer the challenging question here.

A plea to Google

Now, I’m not going to sit here and beg Google to go back to the page of ten blue links. That era is officially dead and for what we get in its place, taking my “protecting the business” hat off, I’m happy it’s gone.

What I will ask for, however, is a more equitable use of it — maybe the memory of it can remember key sites people go to for stories and maintain those referrals. We could even go wild and have a pay-per-referral model to keep the original creativity of the internet alive.

Because while there are certainly ways that publications like ours can adapt (and if you’re not yet, you should join our Tom’s Guide club for free), these can only realistically go so far in making up for the inevitable shortfall of a key site referrer dropping to zero.

For Google to exist, it needs a financially viable, active internet to source from, search across and show its ads on too, which this would surely actively harm. This is a generational leap for search, but one that has some serious problems that need addressing fast.

More from Tom's Guide

• I just tested Xreal’s Android XR glasses, and they are an early glimpse of my dream future smart glasses
• Ask YouTube 'entirely reimagines' how you find videos — here's how it works
• Google just launched Gemini 3.5 Flash — here's all the upgrades

Most Android users will know about QuickShare, the Android equivalent of Apple's AirDrop, but what you might not have known is that there are two ways to send files. The most obvious is to find a nearby device and beam files over via Bluetooth, but you also have the option to connect and send files via QR code instead.

This feature has been part of QuickShare for a while, but it is about to get a much-needed AirDrop-infused boost from Google. The Android-makers just announced that you'll be able to start using those QR codes to share files with iPhones — even if your phone doesn't officially support file-sharing via AirDrop. It's not quite as elegant a solution as traditional QuickShare, but it works, and we can't complain about that.

Here's how to share files via QR code with Android QuickShare

There you have it, an easy way to send files using a QR code instead of the usual QuickShare method. Not only does this mean you can share files with iPhones more easily, it also means you have a backup in case there's any issue with the normal QuickShare linking system.

I can't tell you how many times I've tried to send files between devices, only to find my phone isn't registering that there's another suitable device in range. It's especially problematic in busy areas, when there are a lot of potential devices to connect to.

QR Code sharing means you can set up that connection pretty quickly, and without having to make your devices visible to random strangers who may not have the best intentions.

More From Tom's Guide

• This underrated Motorola Razr Fold feature makes me more productive — Google, Apple, and Samsung should take notice
• I put the iPhone Air vs. iPhone 17e through a 10-round photo shootout — here's the winner
• 9 phones with the longest software support — Samsung, iPhone, Pixel and more

So as you well know by now, the “intelligent eyewear” from Google and Samsung is launching this fall with designs by Gentle Monster and Warby Parker. While nobody was able to go actually hands (or eyes) on with these actual glasses, I did get to test the software that will be fuelling these specs and, well, the Ray-Ban Metas are kind of cooked.

Basically, everything AI-related that Meta has brought to the table in its past couple generations has been usurped by what is set to launch on Google’s glasses. We may not know what they will be called, and we don’t know the price. But if they’re competitive to Meta’s specs, Zuckerberg has a big problem on his hands.

Judging the styles

So we’ve got Warby Parker and Gentle Monster doing the designs, and I wouldn’t be surprised if most people went for the former. We only got to look at photos (ignore publications that said they got to “wear them all,” because they didn’t), so I can only give you my eyes-on perspective.

The Warby Parker frames look less daring and more subtle on your face, whereas Gentle Monster is looking to make a statement. Nothing wrong with that at all, but the aesthetic is not going to be for everyone.

As for specifics on them, Google didn’t share much. But with these prototype specs weighing less than 50 grams, I’m quietly confident that they will also be lightweight and wearable in the very long-term. Plus, they’ll definitely be prescription-friendly, as has been confirmed extensively here at I/O.

All-round smoothness

As for using them day-to-day, you’ll get your standard interface that uses your phone as the brains of your glasses, with a little bit of local AI to run the basics. It’s a method that has a proven track record so far, and they will be compatible with both Android and iOS devices.

Gemini Live is the key to making these true Ray-Ban Meta beaters, and it frankly makes Meta AI seem feeble in comparison. Rather than having to follow rudimentary vocal steps to do certain things, a press and hold on the stem turns it on and I’m having a full conversational interaction with the AI about the world around me.

As is the case with any of these demos, I had a bunch of props to work with — figuring out a Van Gogh painting, or going through a Korean cookbook and ensuring that the recipes are safe for my peanut-allergic friends to consume. In all of these, the AI voice is extra descriptive of its determination and provides a nice helping hand in a way Meta AI has never really done so.

Throw in all the Google app integration in here too and it gets even more interesting. You can receive a summary of your latest messages or turn-by-turn directions to the nearest bar (it’s been a long day, don’t judge me). It’s clear that a real-time AI like Gemini has been the missing link that’s going to bring everything together in a way that can reduce the number of times you need to pull out your phone.

Turning on the display

This is a feature that will not be coming to this first crop of smart glasses — they are all display-free. But the monocular waveguide display in this prototype does add some fantastic visual aids to some key use cases.

You’ve probably already read a lot about this when Mark went eyes-on with this same prototype last year. But to give you my thoughts, bringing Google services into a pair of smart glasses really brings a user-friendly advantage here with the likes of Google Maps and photo taking.

Seeing a little HUD give directions or a GTA-style radar when you look down for a nice street layout is a nice touch. Then the image generation efforts with Nano Banana were frankly hilarious. I took a picture of my camera guy and asked it to turn him into a superhero…what does GL mean? Also his name is Paul Antill as clearly stated on the badge, but now he’s Paul Arnold!?

Regardless, it’s clear that you’ll have plenty of fun with the generative AI experiences of these specs.

Watch out, Meta!

And all of this exposes that while the Ray-Ban Metas may have a couple of years headstart, the usefulness updates over time have been rather minor and slow in the progress of things.

I mean, sure, there are reminders of things, but do they actually mean something when they’re not integrated into your services that matter? Or do turn-by-turn directions really help when the mapping technology doesn’t hold a candle to Google Maps?

So, I say this to Zuck. You have about five months-ish to figure out how to compete with this. I’m not sure what the answer is, but Google is about to bring actually useful AI to smart glasses, and Meta AI needs to catch up fast in both a simple Q&A fashion and in agentic use, too.

More from Tom's Guide

• I lived in the Viture Beast for 10,000 miles, and it just ruined other AR glasses for me
• I just tested Xreal’s Android XR glasses, and they are an early glimpse of my dream future smart glasses
• I switched my Ray-Ban Metas for L’Atitude 52°N for a month, and while there’s cool features for explorers, it’s no contest

I've been using the Ray-Ban Metas, the Meta Ray-Ban Display and the Oakley Meta Vanguard for the better part of a year, and they've proven to be some of the most capable and best smart glasses around. I've found them exceptionally handy for taking pictures and video on the go, listening to music, keeping tabs on conversations, and finding out general info about things that are in front of me.

However, at Google I/O 2026, Google finally revealed Google Glass 2.0 its Intelligent Eyewear, which are due out this fall. Although the introduction was brief, what I saw impressed me in a few ways — most notably, the things they offer that Meta's smart glasses do not.

Fortunately for Meta, it has a few months to improve its own products — which no doubt we'll see at Meta Connect 26 in September. Here are the three big things that Meta will need to add to keep up with Google.

Maps and directions

Apart from search, one of Google's biggest assets is Google Maps, so it's no surprise that turn-by-turn directions are one of the features included in its eyewear. Not only that, but Gemini will also be able to add stops to your route as you go.

Currently, only the Meta Ray-Ban Displays provide native navigation, and it's only recently been updated to include the entire U.S., and a few cities abroad, such as London, Paris, and Rome. And, they're only walking directions.

This is something that's going to have to get a lot more comprehensive, and also become available on all of its smart glasses.

Taking (and editing) photos

More than anything, I use Meta's smart glasses to take photos and videos, as it's so much more convenient and faster than pulling my phone out of my pocket. However, smart glasses' cameras are much less capable than those on smart phones, so I know I'm making a bit of a tradeoff.

Google's smart glasses will let you use Nano Banana to instantly edit the photos you take, to remove background objects or even do something sillier, like putting mustaches on everyone in the picture. Hilarity aside, the Nano Banana integration could help remove, or at least mitigate, one of the key disadvantages of a smart glasses' camera.

A recent update to all of Meta's cameras lets you automatically crop photos where a portion is covered by a hat or hair — a very welcome feature — but it's going to have to go a bit further. Meta does allow you to use AI tools to edit photos you take from within the app, but you can't make these changes proactively.

Greater app support

Thanks to its history of acquisitions, Meta has some pretty useful apps in its portfolio, most notably Whatsapp and Instagram. And, its smart glasses also work well with major music apps (Spotify, Apple Music, Amazon Music) as well as Garmin and Strava. But that's not nearly enough.

At the keynote, Google announced that its glasses would work with a host of apps, such as Uber, Doordash, and more. It's not clear if the apps need to be integrated with the glasses themselves, or if you merely need to connect them via Gemini — which would save a lot of app development time.

Regardless of how it's accomplished — and this is something I noted in my Meta Ray-Ban Display review — there needs to be far greater integration with third-party apps and services.

To its credit, Meta did announce it was launching developer access, so you can start building Web apps for the glasses. But, this will need to be extended to all of its devices, not just the premium Displays, and it will need to get some major companies on board.

Outlook

There's a lot of time between now and when the Google Intelligent Eyewear makes its debut this fall, and Meta does have a strong early lead in actually producing smart glasses.

Still, Google has some real strengths where it can leverage its massive portfolio of businesses, services, and AI to its advantage. Before Thanksgiving rolls around, Zuckerberg and Co. could be playing catch-up — and by then, it may be too late.

More from Tom's Guide

• Google’s new $100 AI Ultra plan just changed the AI race — here's what you get
• Google just launched Gemini 3.5 Flash — here's all the upgrades
• Google unveils Gemini Spark — a '24/7 personal AI agent'

Alongside AI glasses made with Samsung, Gentle Monster and Warby Parker, Google is also tying up with Xreal on Project Aura. We’ve been seeing small glimpses of them for the past year and finally, I was able to go fully hands-on (and eyes-on) at Google I/O. And in three words: it’s the future.

For those uninitiated, Project Aura is Xreal’s best pair of AR glasses connected to a compute puck with Qualcomm silicon inside — taking on Meta’s Project Orion from a while back and figuring out how to stuff the experience of a full-blown VR headset like the Samsung Galaxy XR or Apple Vision Pro into something the size of a slightly oversized pair of specs.

As the smart glasses guy on the team, it’s been my dream to watch the worlds of the best smart glasses and best VR headsets come together. And while they’re definitely not all the way there (you won’t be taking these outside in public without getting second looks), it’s a damn good starting point.

A new window

The field-of-view battle has been ongoing between the likes of Xreal and Viture, and so far that ceiling has been 58 degrees in the Viture Beast. When Xreal’s CEO came out and announced a 70-degree field of view for Project Aura, I was very much in the “I’ll believe it when I see it” camp…

Well, I saw it and yes, it’s a true generational leap for immersion. The prisms are massive and while you can definitely see some edge blurring, the fact that said blurring is kind of moving off to your peripheral vision masks it nicely.

It’s that missing link between just being a nice big screen for your phone or laptop on your face and becoming that stepping stone into being a VR experience to boot. On the sides, you’ve got two cameras along with a centrally-placed lens for full 6 DoF tracking, and the whole thing is tethered to a compute puck (with a strap to shoulder it) that’s roughly the same size as my iPhone 15 Pro.

We don’t know anything about what’s inside the puck except for a 4455mAh battery (probably good for 2-3 hours of heavy use), and some unnamed Snapdragon chip inside.

My guess is an XR2 Gen 2 or something maybe newer (Gen 3? Couldn’t be sure). But the end result is mostly accurate hand tracking and impressively fixed windows in place — to the point that you can walk around them a la VR headsets.

Getting to use them

So what can you do on them? Well, spatial computing stuff of course — this is Android XR after all! If you’ve used the Galaxy XR, you’ll feel right at home with the pinches and drags. But the transparency of glasses adds a new dimension of layering to the whole thing. This is not you seeing the world through cameras, it’s seeing the real world with your content laid over it.

I got to multitask without any noticeable hitches, messed around with a music creation app that merges drawing in the air with Gemini’s music creation, and even played Demeo — a D&D-style tabletop game that almost feels purpose-built for a device like this.

Ultimately, it does hit those same barriers that any spatial computing interface faces: namely that it’s slower to do things than on an actual computer. But I can’t deny that these are so much fun to use! And Gemini brings it all together nicely with voice prompts to make your way around.

Also shout-out to the auto electrochromic dimming of the lenses, which means that if someone is getting my attention, I can look away and the lenses immediately become fully transparent so I can maintain eye contact.

Outlook

OK, so how much are they going to cost and when will they come out? On the latter, they’re set for later this year but everyone’s staying super tight lipped on price. So all I can do is guess, and while I’d love for these to be sub-$1,000, I don’t see a world where this happens. I mean the Galaxy XR headset is nearly two grand, and this is a lot of compute that has been ruthlessly shrunk down into a pair of glasses.

However, doesn’t take away from one simple truth: these are a glimpse into the exciting future of AR glasses — specs that are no longer just a display on your face, but a whole platform to do things on. It’ll take time to get that compute puck outta there, to make them a true must-buy. But as they are now, these are the next generation to look forward to.

More from Tom's Guide

• I switched my Ray-Ban Metas for L’Atitude 52°N for a month, and while there’s cool features for explorers, it’s no contest
• I put Viture Beast and Xreal One Pro head-to-head for a week: Here's the one pair I'm actually keeping in my bag
• I traveled 5,000 miles with Rokid Glasses — this Meta Ray-Ban Display rival impressed me

Google has already thrown its hat into the AI image generation and editing space with its Nano Banana AI image generator powered by Gemini.

Nano Banana, which is now in its second iteration, has been used to help millions of users create professional-grade images through their text descriptions and image prompting, plus make adjustments to those visual creations via reference images that guide the process. Now Google has released another impressive piece of AI-assisted visual generation tech that promises to create videos from all manner of inputs, which include text, images, audio or even videos themselves.

Alongside the ability to create new visuals based on those, you can also edit your newly generated videos through simple or complex conversations with Gemini Omni Flash.

Here’s a deeper look at what Google’s latest AI video generator is capable of and how you can use it.

Generating videos through a variety of inputs and editing them through chats

Gemini Omni Flash is a multimodal AI video generation tool that promises to create videos from any voice references you mention about images, text or audio or just rely on input references. Your voice references can be used to describe the visual language needed to birth your visual creations or you can implement input references such as images of characters, scenes or drawings to apply the styles, motion or effects you want in your video clip.

Google points to Omni Flash as being a highly intelligent AI video creation tool that can build incredibly lifelike scenes and work on its own to produce what happens next in each clip it generates. Its cohesive knowledge of gravity, kinetic energy and fluid dynamics helps it produce more realistic-looking scenes. Omni Flash is also intelligent enough to incorporate Gemini’s understanding of language, imagery and meaning to create short or lengthy explainers from your vocal prompts.

On the video editing front, Omni Flash can listen to your vocal instructions and build upon each one to craft your newly created video any way you see fit. You can speak to the AI video generator to change certain aspects of your video or change everything in it altogether. You can also tell Omni Flash to edit a video you recently shot by altering what’s actually happening in it, adding in new objects or characters or completely transforming a specific moment into something else entirely.

Google is already rolling out the first member of the Omni family to the Gemini app, Google Flow and YouTube Shorts. And as a starter project, you can generate videos with your voice by using Google’s AI “Avatars,” which is a digital version of yourself that uses your voice to sound just like you.

Bottom line

Gemini Omni Flash’s multimodal AI capabilities sound impressive, especially since you’re able to take full advantage of its video generation capabilities by simply voicing your commands. Building worlds, swapping in/out all forms of human and animal characters, switching up visual styles, altering certain details and more come together to offer one of the most powerful AI video generation tools we’ve ever seen.

More from Tom’s Guide

• How to watch Google I/O 2026 and all the announcements to expect
• We're at Google I/O 2026 LIVE — Gemini 3.5, Smart Glasses and all the biggest news
• I let a viral AI agent take over my PC — and now I see why apps are dying

Google has unveiled a new personal AI agent called Gemini Spark at Google I/O today, and it could be the tool that pushes agentic AI into the mainstream. The new tool runs on dedicated virtual machines on Google cloud and is available 24/7 to help "navigate your digital life."

The point of agentic AI is to have an assistant that doesn't just respond to your prompts, but takes action on your behalf. But while opening up your desktop to OpenClaw isn't something everyone is willing to (or should) do, Gemini Spark is different.

For starters, Gemini Spark will be integrated with Google's suite of products like Gmail and Google Docs that billions of people use around the world every day.

The second, and more crucial aspect, is that Google has built Gemini Spark using the Google Antigravity development platform. This is an AI-native IDE (Integrated Development Environment) that prevents the agent from going rogue. So users can effectively send it off to work in the background on tasks that involve agentic coding, knowing that it can be configured to adhere to unbreakable laws.

While Spark will launch initially with Google's first-party tools, in the coming weeks, it will also be able to integrate with third-party tools via MCP (Model Context Protocol). If you're not familiar with MCP, it's the open standard that will allow Spark to operate on your behalf with non-Google services, like GitHub or Notion.

What can Gemini Spark do?

Because Gemini Spark runs in the cloud, it can be accessed at any point from whichever device you're using. You could start interacting with it across Gmail from one of the best Android phones and pick it up later in Google Docs when you get to your laptop. And because it's on at all times, you can set it a task to complete while you're away from your screen — taking a call or getting lunch.

For example, if you use Google's products in your workplace, you could prompt Gemini Spark to pull information from Google Sheets, Docs or Drive and use it to compose an email to a colleague in Gmail, giving them a status update. Or, if you're a student using Gemini Spark on campus, it could work to update your personal study guide in real-time as new assignments arrive from your professor.

Here's a few more actions that, Google says, Gemini Spark will be able to do on your behalf:

• Set recurring tasks or triggers: Automatically parse monthly credit card statements to flag new or hidden subscription fees.
• Teach it new skills: Direct it to check your inbox for ongoing updates from your kids' school
• Create complete workflows: Ask it to synthesize raw meeting notes across emails and chats, create polished Google Docs with its findings and even draft the companion email kicking off a project.

Google says it will be connecting Spark to a ton of third-party services in the future, including services like Doordash, Asana, Dropbox, CapCut and Uber.

While Spark won't launch on Chrome, Google says it will add it to the browser in time. Meanwhile, Android users will also be able to view live updates on tasks assigned to Spark through a dedicated UI space called Android Halo.

When will Gemini Spark be available?

Google says it is taking a phased rollout to Gemini Spark, starting with trusted testers this week. That will follow as a beta for U.S.-based Google Ultra subscribers next week. Unfortunately, we don't yet know when Spark will come to other Google tiers, including AI Pro and AI Plus, as well as free users.

Bottom line

Agentic AI has been on the roadmap for several years now and all of the big tech companies are chasing a viable product. We've already seen it on the periphery of mainstream usage with OpenAI's Operator and Opera's Agentic AI browser. But this feels like it could be a pivotal moment.

Because Google's products are used so widely across the web for both working and creating, the company is expertly poised to introduce an AI agent into tools millions are already familiar with.

I'd bet your garden-variety web user probably doesn't want to install OpenClaw, hand over the keys to their calendar and walk away. But that same user will probably be happy to have Spark reach out to their contacts, agree on a date to connect, find a venue via Google Maps, and make a note in the Google Calendar for them.

Of course, Spark is just one part of Google's overall AI game plan and you can check out everything else the company has announced today on our Google I/O live blog right here.

More from Tom's Guide

• Google I/O 2026 Live: Gemini, Smart Glasses and all the latest news
• I used ChatGPT to build the ultimate 2026 summer bucket list and it totally nailed my vibe
• I tested the viral AI agent that could replace apps — and it made me appreciate my computer without it

Google’s vision for AI is undergoing a massive shift. Today, Google introduced Gemini 3.5 Flash as the latest in their family of models combining frontier intelligence with action. The faster and smart model delivers frontier performance for agents and coding, while excelling at complex long-horizon tasks that deliver real-world utility.

This launch marks a next-generation leap, one that's built entirely around “agentic workflows.” According to Google, the company wants AI to stop behaving like a chatbot you constantly have to manage, and start acting more like an invisible digital layer that can independently complete complex tasks from start to finish.

The next wave of AI at a glance

One thing Google has made clear: Gemini 3.5 is built for action rather than simply conversation. Instead of simply generating text or summarizing information, Gemini 3.5 Flash is designed to execute multi-step workflows, maintain software projects, prepare documents, coordinate “subagents” and handle longer-running tasks with far less human oversight.

According to Google, the model outperforms Gemini 3.1 Pro on several coding and agentic benchmarks, including Terminal-Bench 2.1 and MCP Atlas, while also dramatically reducing response latency. That's incredibly important because AI agents often feel frustratingly slow.

While tools from companies like OpenAI, Anthropic and Meta can already perform surprisingly advanced workflows, many still feel clunky when juggling multiple steps, browsing tools, or reasoning through complex tasks. Google promises that Gemini 3.5 Flash removes that tradeoff between reasoning power and responsiveness.

Google Spark proactively assists with workflows

But today, Gemini Spark may be the even bigger story. The persistent AI agent designed to quietly run in the background, takes actions on your behalf while remaining under your supervision.

According to Google, Spark can proactively assist with workflows, coordinate tasks and function more like an always-on digital layer than a standalone chatbot. Trusted testers will receive access first, with a broader rollout to Google AI Ultra subscribers in the U.S. expected next week.

The industry has swifty moved beyond building the smartest chatbot to building the most capable AI Agent that becomes the invisible operating layer for your digital life. This includes options such as managing schedules, organizing documents, automating repetitive work, coordinating apps and eventually making decisions with minimal prompting.

A dramatic shift from where AI was even a year ago

The first wave of generative AI was about proving models could generate human-like responses. The next phase is about turning those models into autonomous systems capable of executing tasks across your digital environment.

That’s why terms like “agentic AI,” “subagents” and “ambient intelligence” are suddenly appearing everywhere across Silicon Valley.

With AIG, safety concerns are often top of mind. But Google plans to emphasize safety and says Gemini 3.5 was developed under its Frontier Safety Framework, including expanded cyber and CBRN safeguards designed to reduce harmful outputs while minimizing unnecessary refusals on safe prompts.

That balancing act has become increasingly important as AI systems become more autonomous and capable of taking real-world actions. Additionally, the company also confirmed that Gemini 3.5 Pro — the larger and more capable version of the model family — is already being used internally and is expected to launch publicly next month.

If Gemini 3.5 Flash is about proving speed and responsiveness, Gemini 3.5 Pro may ultimately show how far Google believes autonomous AI systems can actually go.

Bottom line

Gemini 3.5 Flash is generally available today via the Gemini app and AI Mode in Google Search.

With this launch, the company is betting that the future of AI isn’t a tool you constantly open and prompt, but an always-present system quietly managing the chaos of digital life in the background.

And after years of AI being framed as a smarter search box, that may be the biggest shift yet.

More from Tom’s Guide

• I sold band merch for some of the biggest names in music — here’s why AI can’t replace real artists
• I could parent without AI — but I could also churn my own butter
• I use the ChatGPT ‘fart’ prompt whenever I’m stuck — and it somehow works every time

Today (May 19), during Google I/O 2026, CEO Sundar Pichai announced new features that will change how you search for videos on YouTube.

Ask YouTube is a new way to search for videos in the app, which Pichai said "entirely reimagines the experience." The AI feature lets you ask more complex and conversational search queries in the search bar.

Plus, Gemini Omni is getting deeper integration in the YouTube Shorts Remix and YouTube Create tools. Here's what's new.

Google's example was looking for tips on teaching your kids how to ride a bike or finding reviews of cozy games to play before bed.

Ask YouTube will compile videos that best match the search, which includes both long-form videos and Shorts. Similar to AI overviews in Google Search, you will see an "interactive, structured" response. In this new view, Ask YouTube will jump you directly to the part of the video that relates to your query.

The potential impact on YouTube Creators seems like it could be pretty large, as they constantly battle the algorithm. My colleague Marcus Cooper expressed skepticism about the new feature.

"On the surface, Ask YouTube sounds like it could be helpful to the user. Whether it will help people find more varied creators than the algorithm — we'll see," Cooper said. "Do we need everything to be AI? Probably not."

Initially, the feature is only be available to YouTube Premium subscribers aged 18 and up in the United States, starting today. A broader release will roll out later, but Google did not provide a timeline.

Gemini in Shorts and Create

If you make and upload videos to YouTube, the new Gemini Omni tool is coming to Shorts Remix and the YouTube Create app.

Remix allows you to take an "eligible Short" and add your own AI prompts and images to create a new vision, like adding a 90s filter or inserting yourself within the context of the original video. Gemini Omni is supposed to better understand "user intent" for more consistent storytelling while handling complex video and audio adjustments.

Remixed Shorts created via Omni will have digital watermarks, identifying metadata and will automatically link back to the original video, Google said. Creators will have the option to opt-out of the remix.

Remixing with Omni is rolling out today for free in YouTube Shorts Remix and the Create app. It will also come to AI Playground in the near future.

More from Tom's Guide

• This underrated Motorola Razr Fold feature makes me more productive — Google, Apple, and Samsung should take notice
• Stop letting Google track your every move — how to clear your search history in 2026
• Google settles Android class action lawsuit for $135 million with payouts to 100 million users — here's how to claim your share

Considering Google I/O is a developer event that mainly focuses on software, it's rare to see any big hardware announcements. But that's exactly what's happened at this year's conference, with Samsung and Google offering a sneak peek at their new smart glasses — or "intelligent eyewear," as they've both put it.

The duo has unveiled two different designs, created with Gentle Monster and Warby Parker. These aren't AR Glasses that are designed to run on Android XR, though. They're more like the Meta Ray-Ban glasses, and are designed as a way to interact with your phone and AI with your voice.

Samsung describes the glasses as being "built to understand the world alongside users in real time," with the goal being to keep users connected while staying "hands-free and heads-up." In other words, it's a way of interacting with your phone without actually having to have it out of your pocket or bag — aided by Gemini.

Features mentioned by Samsung include notification summaries, the ability to add events to your calendar, navigation assistance, personalized recommendations about locations on your walking route, and even real-time translation. The translation sounds particularly interesting, because Samsung says it will be able to match the translated audio to the speaker's own voice, while also translating text on any signs or menus that you're looking at.

And yes, these glasses do have a pair of camera lenses on the front. That helps power various AI features and will offer the ability to take photos without your phone.

Samsung has made it clear that the Intelligent Eyewear isn't just about being able to sport technology on your face. It's about expanding the Galaxy device lineup in a way that merges fashion and technology. This is no doubt why Samsung is releasing multiple designs, in collaboration with other eyewear brands.

The first collections will launch this fall in "select markets," though Samsung doesn't offer any additional details beyond that. Presumably, we'll be hearing more at Galaxy Unpacked in July, ahead of a fall release date.

More from Tom’s Guide

• Google I/O 2026 LIVE — Android 17, Googlebook, Android XR Glasses, Gemini Intelligence and more
• Stop letting Google track your every move — how to clear your search history in 2026
• Apple WWDC 2026 keynote announced for June 8 — and the invite could be the new Siri

At Google I/O 2026, the company unveiled a massive expansion of its AI subscription strategy — including a brand-new $100-per-month Google AI Ultra tier built for developers, creators and power users.

With this new plan, Google is leaning hard into AGI or "agentic workflows” — AI systems capable of planning, coordinating and completing tasks with far less human involvement.

And at the center of that strategy is something called Gemini Spark, a model Google describes as a “24/7 AI agent.” The model can connect information across your Google ecosystem, take action on your behalf and help manage your digital life under your direction.

That’s a major evolution from the chatbot era we’ve previously experienced with Gemini.

What you get with AI Ultra ($100/month)

Google’s new $100 AI Ultra plan targets developers, technical professionals and advanced creators who need significantly higher usage limits and more advanced tools.

Everything in Pro, plus:

• 5X higher usage limits than Google’s Pro plan
• Gemini 3.5 Flash integration
• Early access to Gemini Spark, Google’s new 24/7 AI agent system
• Faster workflows using Gemini 3.5 Flash for testing, debugging and coding
• 20TB of cloud storage
• YouTube Premium included

As the AI subscription market is quietly becoming one of the biggest battlefields in tech, Google’s approach appears to be leaning heavily into ecosystem integration, which is their advantage here.

Unlike many competitors, Google already owns the products people spend most of their digital lives inside: Gmail, Calendar, Docs, Drive, Android, Chrome and YouTube. Gemini Spark suddenly makes much more sense when viewed through that lens.

Other Google AI subscription tiers explained

Google now has multiple AI subscription levels, and they’re increasingly designed around different types of users — from casual Gemini users to developers and power users running advanced AI workflows.

AI Plus ($7.99/month)

This is Google’s mainstream paid AI tier for everyday users who want access to the latest Gemini models and productivity tools.

What you get:

• Access to Gemini Omni, Google’s multimodal AI model for text, image and video creation/editing
• Access to Gemini 3.5 Flash for faster AI performance and coding tasks
• AI Inbox in Gmail for smart email triage and AI-generated replies (U.S. only)
• Daily Brief AI summaries pulling from Gmail, Calendar and Gemini chats (U.S. only)
• Standard AI usage limits based on Google’s new “compute-used” system

AI Pro ($20/month)

AI Pro is aimed at heavier productivity users and creators who want more advanced tools and additional bundled perks.

Everything in AI Plus, plus:

• Higher AI usage limits
• Additional voice features in Gmail, Docs and Keep coming later this summer
• Access to Google Pics, a new AI image generation and editing tool
• YouTube Premium Lite in select countries at no extra cost
• Access to optional pay-as-you-go AI credits for additional compute power in Google Antigravity, Google Flow and Gemini
• Included Health Premium and Home Premium benefits

AI Ultra ($200/month)

Google also lowered the price of its flagship Ultra plan from $250 to $200 while keeping the same feature set.

Everything in the $100 Ultra plan, but with even larger limits:

• 20X higher usage limits than Pro
• Full access to Google’s most advanced AI systems
• Expanded access to Project Genie, Google’s experimental AI world-building platform
• New Street View-powered world generation tools inside Project Genie

Google says Gemini 3.5 Flash delivers frontier-level performance at dramatically faster speeds, something developers increasingly care about as AI tools evolve from novelty demos into real workflow infrastructure.

The company also announced several new AI-powered productivity tools that push Gmail further toward becoming an AI workspace hub rather than just an email client.

In other words, Google increasingly sees Gemini as the intelligence layer sitting across your entire digital life.

Final thoughts

Google is betting that AI assistants become infrastructure, which is one of the big takeaways from Google I/O 2026 today. It's clear that Google is committed to the idea that AI assistants will evolve into consistent workflow systems that actively manage information and decision-making in the background.

What a difference a few years makes. The basic chatbots we started with feel like ancient history now, and Gemini Spark is proof that the rivalry between Google, OpenAI, and Anthropic is only heating up.

More from Tom’s Guide

• I sold band merch for some of the biggest names in music — here’s why AI can’t replace real artists
• I used NotebookLM to make a 'Family Wiki'— and now everything I need to run the household is a click away
• I use the ChatGPT ‘fart’ prompt whenever I’m stuck — and it somehow works every time

I am constantly finding new uses for NotebookLM. Now that it's available in the Gemini app, it's been even more helpful when I need answers on-the-go. I recently discovered a use for it that has made my life as a mom of three significantly easier. Let me explain.

Like most people, I juggle a lot. And, like most people, I try to stay organized with digital support such as Alexa+ and Google calendar. Those are helpful beyond words, but when I'm looking for a specific pediatrician's note, the warranty for the dishwasher, my grandmother's apple cobbler recipe or my daughter's IEP report from last year, I need to be able to pull those up quickly.

Sure, Gemini can search my inbox and look through my Google Drive. It can even sift through my Photos and screenshots, but that's not always enough. So, I started experimenting with using NotebookLM as a personal “family wiki” — and honestly, it may be one of the most underrated AI tools for normal people right now.

Everything has a place

Most people think of AI as something that searches the internet for answers. Although NotebookLM can do that, it also works by scraping the documents you upload. That's why I now keep everything from my household manuals and recipes to report cards and perscriptions in NotebookLM.

Now, I can use AI to search for things like, "IEP notes from Fall 2025." I can get the entire report from the school or just the details in bullet points. I can instantly compare my kids' grades from one report to the next or learn when the warranty on our fridge expires.

Because all of the answers are filed entirely on your actual life, your documents, your memories and your history, NotebookLM builds a searchable external hard drive for your brain.

The freedom of finding information immediately while on a call to the doctor, the school or even the phone company is life changing. I feel more organized than I have ever been in my life.

What makes NotebookLM different?

Unlike traditional AI assistants, NotebookLM is grounded in your own sources. I have uploaded, PDFs, Google Docs, notes, spreadsheets, receipts, cods, medical summaries, travel iteneraries, manuals and schedules. I simply scan or take a photo with my phone. From there, I get responses from NotebookLM such as

“According to the documents you uploaded, your son’s allergy symptoms first appeared in March 2024.”

That’s a massive shift. I would never have rememered that. But NotebookLM lets you turn just about anything into a search engine of sorts. It's like a personalized memorty system.

It transforms AI from a search engine into something closer to a personalized memory system. So, when I realized I could organized my entire family this way, I stopped thinking of it as just a research tool.

Instead of one giant messy notebook, I created separate notebooks for different areas of life.

Each family member has their own notebook

It's hard to remember everything, which is why making a family "wiki" of each family member has really helped. I've found it especially helpful with medical information. My daughter has special needs and her medication is always changing. Remember all the perscription names, when each one started and when each was stopped is a lot to keep track of — now NotebookLM helps.

It is a huge support for medication histories, vaccination dates, reccuring symptoms, specialist recommendations, old lab work, doctor visits and paperwork. Now, I can pull up whatever I need in a moment's notice. I simply ask:

“When did the knee pain first appear in the records?”

“Compare cholesterol levels between 2024 and now.”

“Which antibiotics caused side effects previously?”

Instead of hunting through portals and PDFs, the information becomes conversational.

NotebookLM is always a huge help for homeowner survival. Every homeowner has a junk drawer full of manuals nobody can find. Now I just use NotebookLM to retieve appliance manuals, paint color codes, mortgage documents, warranty information, repair receipts and contractor invoices

I simply ask:

“What water filter does the refrigerator use?”

“What paint color is in the upstairs bathroom?”

“Who repaired the water heater last year?”

NotebookLM may secretly be one of the best organization tools available

I've created notebooks for AI research, fitness studies, writing courses, my kids' homework and art (I can't keep all the physcial stuff!). Then I ask it to summarize or generate information based on each notebook. If I'm looking for "that adorable image of a goldfish" my kid drew in preschool, I can easily find it.

And, because it only uses your materials, the responses stay focused instead of drifting into random internet noise.

That said, there are two things I've learned very quickly.

• Keep notebooks separated. Even if you're just starting out and you haven't uploaded a lot of documents yet, trust me on this, you're going to want to keep individual notebooks. That way the AI won't get medical data confused with different family members. You'll stay far more organized than you even realize if you do this from the start.
• Treat privacy seriously. Google says NotebookLM data stays private and isn’t used to train public AI models. But basic digital hygiene still matters. So you're going to want to ensure you remove Social Security numbers, banking information and avoid ultra sensitive documents.

Bottom line

I really should have thought of using NotebookLM to create family notebooks sooner. For the first time, my personal history, household knowledge, work achievements and family systems are searchable in a convenient way. I can ask NotebookLM the questions I'm getting from the school nurse or telephone company, and pull up the information immediately.

And once you experience that, it’s hard to go back to digging through folders ever again.

More from Tom’s Guide

• Forget bank accounts — why you should be more terrified of a ChatGPT chat history leak
• I asked ChatGPT to apply Taylor Swift’s ‘Eras’ mindset to my work — and it broke a brutal creative slump
• Google’s new $100 AI Ultra plan just changed the AI race — here's what you get

The days when you could feasibly upgrade your phone every couple of years are long gone. Not only are the best phones more expensive than ever, but the state of the world and the electronics industry mean the cost of buying a new phone is not going to go down anytime soon. Even the best cheap phones seem to be slowly creeping up in price.

With that in mind, it's no surprise that people have been keeping hold of their phones for longer. But if you're going to hang on to your phone for longer, you'll want to make sure your investment is well-spent.

In days gone by, Android phones were typically limited to two years of OS updates and an additional year of security support. The loss of support didn't magically stop your phone from working, but it does mean your phone was more at risk.

Thankfully, phone makers now offer much longer-term support than they did before. Not only would that extend your phone's operational lifespan, but it might also increase its resale value if you ever do decide to swap it out for a newer device.

Here are the nine phone makers that offer the longest support.

Fairphone 6 — 7 years Android, 8 years security (2025 to 2033)

If you're looking for the longest-guaranteed software support from a phone maker, then Fairphone is probably the company you'll want to go with. Fairphone's whole ethos is to sell you a phone that you'll keep for a long time, and that's manifested in easy-to-repair phones and long-lasting software support.

Fairphone has promised that the Fairphone 6 will receive eight years of guaranteed software support. Since the phone was released in June 2025, it means there will be frequent security updates until at least mid-2033. That includes at least seven Android version updates as well, taking the phone from Android 15 to Android 22.

Note that Fairphone has said "at least" for both those pledges. It has previously said that it will aim to keep software support going for longer, if it can, but there's no hard guarantee right now.

Google Pixel 10 — 7 years (2025 to 2032)

Google was the first Android phone maker to promise seven years of software support following the launch of Pixel 8, and it's maintained that pledge for every subsequent Pixel release.

Since the Pixel 10 series was released last August, phones will lose software support ahead of the Galaxy S26 series. That said, since Google still offers the full seven years of support for its mid-range phones, the slightly newer Pixel 10a should still be supported until sometime in 2033 — as should the upcoming Pixel 11 series.

Google used to have a calendar telling everyone the exact date Pixel phones would lose software support. But since altering its policies, we can only guess as to when the phones will be dropped.

In the past, Google confirmed that its increased software support also means spare parts will be available for the duration of the phone's lifespan. Should your phone break at some point in the next seven years, you should still be able to buy the parts to fix it. We haven't heard that promise for a while, so it's unclear whether that's still Google policy, but it is something to think about.

Samsung Galaxy S26 — 7 years (2026 to 2033)

Back in 2022, the promise of four years of Android updates and five years of security updates was considered a big deal, but when Google pledged seven years of support the following year, Samsung was quick to follow suit.

Every Samsung flagship since the Galaxy S24 will get seven years of software support, and for the Galaxy S26, that means those updates will be coming until at least 2033. Most importantly, Samsung has clarified that this seven-year pledge means the S26 will get seven years of security patches and seven major Android upgrades.

That means the S26 will eventually be able to run Android 23, even if the software isn't released until after the phone's seventh birthday is over. It's a similar story with the Galaxy Z Fold 7 and Z Flip 7, though the fact that they're a few months older than the S26 series may mean they miss out on Android 23.

iPhones — Typically 6-7 years iOS updates, variable for security

Apple doesn't announce how long an iPhone will be eligible for software support. We don't actually find out until a new piece of software is announced at the WWDC conference, and even then, Apple hides that news in the fine print.

Each new iPhone that gets released can usually expect to get six or seven years of full software updates, which includes an upgrade to the latest version of iOS and all the new features that come with it. By this estimate, the iPhone 17 series will likely continue to get full iOS updates until 2031 or 2032

Apple is a lot more flexible when it comes to security updates, and will continue to release patches and other protective updates long after a phone has ceased being eligible for new versions of iOS. In fact Apple rolled out a security update for iOS 15 on May 11 2026, which covers phones as old as the iPhone 6s — which was first released almost 11 years ago.

If you think that's generous, iOS 12 also got an update back in January, which covers phones up to (and including) 2013's iPhone 5s. Admittedly, that is the first update those phones have received for three years, but it shows that Apple is still willing to fix certain security problems on older iPhones that may still be in use — even if it isn't willing to make hard promises ahead of time.

Honor Magic 8 Pro — 7 years (2025 to 2032)

Chinese brands aren't officially on sale in the U.S., which makes them a lot harder to find in the states. But if you do have the opportunity to get hold of an Honor Magic 8 Pro you can be happy with the knowledge that it will come with seven years of software support — including a guaranteed seven major Android upgrades.

Chinese phone brands don't have the best history when it comes to software support. But things do seem to be changing, and a bunch of recent releases are offering more than the typical three to four years of support. So don't expect the Magic 8 Pro to be standing alone on this list for much longer.

Motorola Signature — 7 years (2026 to 2033)

Motorola is best known for foldables and cheaper devices, but the newly released Motorola Signature is its attempt to cater to a more premium market. With that in mind, the company has also promised seven years of software support, including security and major Android updates.

Of course, this does come with some caveats. The first is that the Signature is not on sale in the U.S., and the cheaper Motorola handsets don't get nearly as much support. The Razr Ultra 2026 only gets three years of Android updates and four years of security support, for example. On top of that, Motorola doesn't have a positive reputation when it comes to updates, and the company has been criticized for not rolling them out fast enough.

So while the Signature does offer a lot more support, be warned that there may be delays actually getting those updates installed.

Samsung Galaxy A57 — 6 years (2026 to 2032)

Unlike Google, Samsung's software generosity doesn't extend to its mid-range phones, with the newly released Galaxy A57 only getting six years of official software support. Samsung has confirmed that this will include six numbered Android updates, meaning the phone will receive an upgrade to Android 22 when the time comes, as well as six calendar years of security updates.

It's not quite as much support as the flagship devices, but you still have to remember that the Galaxy A57 is a cheaper device. The starting price for the phone is $540, a $250 discount compared to the $799 S26.

You don't get as many premium features on the Galaxy A57, though that doesn't stop it from being one of the best cheap phones. But if you look at things purely from the point of view of software support, it is a slightly better value for those of you hoping to keep your phone for the long haul.

Nothing Phone 4a Pro — 3 years Android, 6 years security (2026 to 2032)

When it comes to Nothing's latest mid-range phone line-up, you're not getting nearly as much support as you might with a flagship phone. The Nothing Phone 4a Pro does a little better than some past Nothing Phones, but it's still pretty limited compared to what's out there.

Three guaranteed major Android updates isn't a lot, and it means the number of new features you'll get to enjoy on the Nothing Phone 4a Pro is quite limited. However, the promise of six years of security updates means you will be able to keep your phone for a significant amount of time before you're putting yourself at risk.

OnePlus 15 — 4 years Android, 6 years security (2025 to 2031)

OnePlus offers a little more than Nothing in the Android update department, but considering the relative age of the OnePlus 15, you're going to run out of new Android updates slightly sooner. Still, four major updates will take you up to Android 20, and the extra two years of security updates will offer protection until the year 2031.

In the past OnePlus has said it doesn't see the benefit of offering several years of full software support, because your phone will be out of date before the deadline. However, I say that misses the point of offering extended software support, especially when it's paired with improved repairability.

Sadly, it seems OnePlus won't get the opportunity to improve on software support in the future, since there are reports that it will be pulling out of certain regions later this year. Updates should still roll out as promised, but it means the OnePlus 15 may be the last flagship that's easily available to buy.

Benefits of long-lasting software support

If you're wondering why you need a phone with long-lasting support, let me explain. The main reason is that several years of consistent security updates ensure your phone stays secure and safe from anyone who might want to break in. Hackers of all kinds are always looking to find exploits and security holes in mobile software.

While the manufacturer is on the hook for software updates, you still have some level of protection from any exploits that might be uncovered. But once those updates stop coming, you're suddenly putting yourself and your data at risk. Having more years of software support means you can keep hold of your older phone for longer, without it becoming a liability.

Similarly, the more years of full software support you get, the longer you'll get access to all the latest features and software upgrades. It doesn't matter whether you have an Android or an iPhone, the longer you keep getting new features, the less likely you are to be tempted by a newer device. It may also have a positive impact on your resale value, since the next owner will be able to take advantage of the phone's longevity as well.

The obvious downside is your phone's hardware may outlast its software — especially the battery. That means there will no doubt be a point where you have to get some sort of repair done on your phone, whether you plan on doing it yourself or via a professional repair shop.

More From Tom's Guide

• Google is secretly phasing out Android apps — and AI is the reason why
• Forget voicemail, Pixel phones have a handy AI feature that takes messages for you — here's how it works
• Forget the iPhone 20 — here's what's coming after the 'four-edge bending' display

There are plenty of features hiding beneath the surface of your phone, and chances are you didn't even know most of them existed. Pixel phones are some of the worst offenders for this, with Google continually adding features and updates to the phones and seemingly expecting users to figure everything out by themselves.

Take a Message is one such feature, nestled in the depths of the Pixel Phone app, and essentially operating as a more advanced version of voicemail.

Much like voicemail, callers will be given the chance to leave you a message, which your Pixel will record and transcribe. Take a Message is an AI-powered feature that does everything on your device, which means it's more private and not subject to the whim of a faceless phone company. Once active, the feature will trigger whenever you miss or decline a phone call.

The best part is that this is all done in real time, meaning you can pick up the phone at any point and take the call if it turns out that it's important. Plus, it isn't exclusive to the latest Pixel 10 series phones and is available on Pixel devices as old as the Pixel 6 and Pixel Watch 2.

There are a couple of downsides, though. The first is that it's limited by location, and is currently only available in the U.S., U.K., Australia, Ireland and Canada — or in other words, the nations where English is most widely spoken.

It's also only available if you have a cell signal and your phone is switched on. If you're in a black spot or your phone is turned off, then your regular voicemail service will handle things.

Here's how to turn on Take a Message on your Pixel phone.

There you have it. Take a Message is now switched on, and it means people no longer have to leave you voicemail messages the way they used to. Instead, it's all handled by an on-device AI that saves a recording of the call, transcribes the contents and gives you the option to listen or share that message to your heart's content.

Realistically, the functional difference between Take a Message and a visual voicemail service is minimal. If your carrier supports visual voicemail, and you have it set up already, you won't notice a whole lot of difference — and neither will the people calling you.

If visual voicemail has never been an option, then you're going to get all those benefits regardless of whether your carrier says you're allowed them or not.

This means that all Pixel owners get to enjoy a consistent voicemail experience, and one that keeps your messages a lot more secure than a normal voicemail inbox might. Hopefully, this feature expands to all Android phones in the future, or at the very least, those devices that don't offer their own take on Take a message already.

More From Tom's Guide

• 'Two-day battery life is a game changer': Honor Magic8 Pro's silicon-carbon battery is a wake-up call for Samsung and Apple
• Google is bringing AirDrop support to more Android phones — here are all the devices that will get it next
• Samsung Galaxy Z Fold 8 just tipped for July 22 launch at Unpacked — and it could be joined by Galaxy Glasses