The use of AI writing tools is everywhere from LinkedIn to Substack and just about everywhere in between. So it's no surprise that AI users are looking for ways to disguise their writing in hopes to bypass AI detectors like GPTZero and Pangram.
To see if these humanizer actually work, I tried several different ones to see the results. As it turns out, the writing ended up sounding more like AI slop than what the AI writing tool churned out in the first place.
Clear sentences came back stuffed with bizarre synonyms and the rhythm became choppy and unnatural. In some cases, the tools changed facts or stripped away important context. The results sounded anything but human.
What AI humanizers actually change
AI detectors look for patterns that frequently appear in machine-generated text. These can include predictable word choices, repetitive sentence structures and a suspiciously consistent rhythm.
Many humanizers appear to target those patterns and then replace common words with less predictable alternatives. These humanizers also chop up sentences and rearrange clauses in an effort to spin whatever AI has written. Frankly, the results are so off that you're better off just writing that essay or presentation yourself just to save yourself the embarrassment of coming back from an AI humanizer.
And yes, these tools do make the writing statistically less predictable, but unpredictable writing isn’t necessarily good writing. Sometimes it’s simply harder to read.
A sentence such as “This phone has excellent battery life” might become “The mobile handset boasts an endurance capacity that remains notably elevated.” Technically, the basic idea is still there. Nobody would willingly write it that way.
1. They sound like bad machine translations
For this experiment, I had ChatGPT write a short essay about the domestication of dogs. The intro read like this:
Dogs were domesticated thousands of years before humans began keeping livestock or growing crops, making them our oldest animal companions. While scientists still debate exactly when and where the relationship began, genetic and archaeological evidence suggests that dogs descended from an ancient population of wolves at least 15,000 years ago. Some studies place the beginning of domestication much earlier, possibly 30,000 years ago or more.
And while that sounds pretty good, it would easily get picked up as "100% AI written" if it was put through an AI detector.
The first humanizer I tried was Superhuman from Grammarly. I'm a huge fan of Grammarly and I use it all day to catch my spelling errors and grammar mistakes. Having never tried the Humanizer, I ran the intro through. Here's the result:
Dogs became our companions long before people started raising livestock or planting crops, making them the oldest animals to live alongside humans. Scientists are still unsure about the exact time and place this relationship began, but genetic and archaeological evidence shows that dogs came from an ancient group of wolves at least 15,000 years ago. Some research suggests that domestication may have started even earlier, possibly over 30,000 years ago
Not bad, but the biggest problem is the language itself. Straightforward words seem regularly replaced with longer, stranger alternatives that didn’t fit the sentence. It just seems as if the Humanizer is too determined to avoid obvious word choices that it removed any trace of a natural voice. Frankly, even the original AI draft sounded more human.
2. The re-wording can actually change the facts
Awkward wording is annoying, but factual drift is worse. When I tried using Rehumanize.io this is the result:
Dogs were around long before we started farming or raising other animals for food, making them our first animal friends. Scientists are still figuring out the exact time and place this connection started. However, based on genetics and old discoveries, it looks like dogs came from an ancient wolf group centuries ago. Some research even suggests domestication could have begun 30,000 years ago or even earlier.
We went from 15,000 years to "centuries." Yikes! That rewrite doesn’t make the explanation more natural. It makes is far less accurate. I saw the same problem with benchmark scores and version numbers when I ran the test on more technical content. The problem is, details like data are not interchangeable the way certain words are. Using a humanizer for legal, medical and safety reports could result in vague generalities.
Once that happens, you can’t trust the output without checking it against the original line by line. Any time the tool might save quickly disappears.
3. They don’t guarantee that you’ll beat an AI detector
Humanizers are often marketed as a way to bypass AI detection, but that promise is shaky. Detection tools change, and they can learn to recognize the same strange synonym swaps and mangled sentence patterns that humanizers produce. During several of my tests, processing a draft through a humanizer actually increased its probability of being flagged.
Also, annoyingly, some tools like Ace.ai wouldn't give me the results until I downloaded an entire browser. And even though I could adjust the level of humanization, it still felt...off. Here is the result of Ace.ai with the same text:
Dogs have been by our side since long before we started farming or raising livestock, making them our oldest and most loyal friends. While the experts are still debating the specifics, evidence from both genetics and archaeology points to dogs descending from ancient wolves at least 15,000 years ago. Some research even suggests our bond might go back much further—possibly 30,000 years or more.
Interestingly, Pangram marked that as AI-written even though it was technically "humanized." And while AI detectors are hardly perfect, rewriting a paragraph to purely satisfy one of them feels like trying to hit an unreliable target while steadily making the copy worse.
What to do instead of using a humanizer
Using AI tools to support your writing isn't something to be ashamed of. It's been a while since I've had to pull an all-nighter in college, but I can only imagine how tempting it could be.
Instead though, you should use AI to help you brainstorm or outline what you're writing. Sometimes the worst part is just getting started. Once you've got an outline, start writing (you, not the AI). Never give AI the first pass of a draft. It's better to use a voice recorder to get your thoughts down and have AI edit your grammar and fact-check.
Using a Claude Skill or Gemini Gem for fact-checking can be a much-needed time saver at the eleventh hour. There are plenty of other ways AI can save you time, especially now that Google is offering Pro plans to students. Using AI for research, outlining and fact-checking can give you the time back you need to do the actual writing — no humanizer required.
Have you tried using a humanizer? Let me know what you thought of the results in the comments.
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. Finally, you can visit our dedicated Tom's Guide Savings Squad hub for expert help on getting the best products for less.
More from Tom's Guide
