As a ChatGPT power user, one of my biggest complaints is that it doesn't always give me an answer in the way I want. For example, I might ask it to edit a paragraph and remove all serial commas or for feedback on an idea. But once it responds to what I thought was a straightforward command, it seems to have gotten confused. Then, I'm suddenly given a five-part explainer that I never asked for. Or, I've recently asked ChatGPT to help with something and it's just responded with "Yes."
If this happens to you, stop trying to solve the issue by adding more instructions and prompts. Instead, try doing something much simpler by adding a slash.
When you type a slash command such as /human and /critic at the beginning of your prompt, you'll get a response that's much more accurate to what you're actually looking for.
To be clear, these ChatGPT commands aren't the secret to unlocking a hidden OpenAI setting. In fact, you can use them with any chatbot including Gemini, Claude or Muse.
Here are the 10 I've found most useful.
1. /human
This is a command you've probably seen a lot lately. It's a viral command that helps change the tone of AI-generated writing.
Try this prompt: /human Rewrite this so it sounds like something a person would actually say: [paste text]
I use it to get rid of canned transitions, excessive explanation, repetitive sentence structures and the strangely polished tone AI tends to fall into before sending an email or drafting a presentation. This is especially useful if you write a draft and then use AI to edit. All too often the chatbot will completely change what I've written. By asking it to "/human only make minor edits on my human-written draft," you should see much better results.
For more control, you can define /human in your ChatGPT instructions with the specific habits you want it to avoid.
2. /critic
Chatbots can be annoyingly nice and people pleasing. Give ChatGPT an idea and it will often find something encouraging to say about it before getting anywhere near the problems. That's useful when you want support. It's less useful when you're trying to decide whether an idea is actually good.
That's when I use /critic.
Try this prompt: /critic I'm thinking about buying an expensive espresso machine even though I only make coffee twice a week. Assess the idea.
The point is to signal that you're not looking for validation. You want weaknesses, risks and things you haven't considered.
It's particularly useful for pitches, plans, purchases and anything you've become a little too attached to.
3. /tldr
Sometimes I don't need the background. I need the answer.
Try this prompt: /tldr Explain what's actually changing in this announcement: [paste text]
The familiar internet shorthand already carries a lot of meaning, so ChatGPT generally understands exactly what you're asking for: compress this.
I find this especially useful for long emails, press releases, documentation and articles where I want the takeaway before deciding whether the rest is worth reading.
4. /senior
There are also times when ChatGPT explains far too much.
If you already know a subject, getting three paragraphs of background before the useful part can be maddening. /senior is my shorthand for skipping the 101 material.
Try this prompt: /senior Review this AI agent architecture. Assume I understand LLMs, tool calling, MCP and context windows: [paste details]
This one benefits from a little context because "senior" can mean different things depending on what you're discussing. Tell ChatGPT the expertise it should assume, and it can jump much closer to the level you need.
5. /diff
This is one of my favorites for editing. If I give ChatGPT 800 words and ask it to make three changes, I don't just want another 800 words back. I want to know what changed.
Try this prompt: /diff Tighten this paragraph without changing my argument: [paste text]
You can also be more explicit: /diff Show me only the sentences you would change and the replacements.
It's useful for writing, but the same idea works with code, prompts and other documents where comparing two nearly identical versions is more work than the edit itself.
6. /bullets
ChatGPT loves prose. Sometimes prose is exactly what I don't want.
Try this prompt: /bullets Give me the biggest reasons not to buy a new phone if mine is only two years old.
This is a tiny command, but it immediately tells the model what kind of answer will be easiest for me to scan.
I use it for meeting prep, pros and cons, shopping decisions, brainstorming and any situation where I'm looking for information rather than a beautifully constructed explanation.
7. /table
This command does the same thing for comparisons.
Try this prompt: /table Compare ChatGPT, Claude and Gemini for writing, research, coding and everyday use.
Instead of asking ChatGPT to "put this into a table with columns for..." every time, /table gives it the general idea.
You'll still want to specify the categories of course, but for quick comparisons this can turn a messy wall of text into something you can understand at a glance.
8. /devilsadvocate
This is the command I use when I think I've already made up my mind.
Try this prompt: /devilsadvocate I think cancelling all my streaming subscriptions and switching between one service each month will save me money. What's wrong with this plan?
Rather than asking ChatGPT whether it agrees, you're explicitly asking it to construct the strongest case against you.
That makes it useful for decisions where confirmation bias is the real problem. If the counterargument doesn't change your mind, you can at least make the decision knowing you've considered the other side.
9. /rubberduck
This one comes from the programming idea of rubber duck debugging: explaining a problem out loud can help you discover what's actually wrong.
You can make ChatGPT play the duck.
Try this prompt: /rubberduck Don't solve this for me yet. Ask me one question at a time to help me figure out why my morning routine keeps taking too long.
I like this because it changes ChatGPT from answer machine to thinking partner, which is always great to avoid brain rot from using AI too much.
Instead of immediately generating a 10-step solution based on incomplete information, you're telling it to interrogate the problem first. That can be useful for debugging code, working through a decision, planning a project or figuring out why something you're doing isn't working.
Try making your own slash commands
This is a useful trick that a lot of people don't know they can do. Because, maybe you like a few of mine, but want commands tailored to you.
You can create /short for answers under 100 words, /pitch for turning an idea into a specific format you use at work, /kids for explaining something to a child or /verify for forcing an answer to separate established facts from assumptions.
The slash itself isn't magic. It's just the easiest way to create a recognizable trigger. For commands where you want consistent behavior, define them in ChatGPT's instructions. You could add something like:
When I begin a prompt with /critic, skip praise and focus on weaknesses, factual vulnerabilities and missing information. When I use /human, rewrite the text with natural sentence variation and remove canned AI language and unnecessary transitions.
After that, /critic isn't merely something you hope ChatGPT interprets correctly. You've told it exactly what the shortcut means.
And that's why I've stuck with slash commands. One character and a word gets me there faster.
Follow Tom's Guide on Google News and add us as a preferred source to get our up-to-date news, analysis, and reviews in your feeds. Subscribe to Tom's Guide on YouTube and follow us on TikTok.
More from Tom's Guide
