checkifmytextisai.com

Why this is different

Other AI text detectors goes the machine learning approach and trains a classifier model on a dataset of human-written and AI-written text. The thing is though, there is no structural difference between human vs AI writing.

For example, while em dashes are a lot more common with AI outputs, it is not a clear or reliable indiciation. The only reason it writes like that is due to the original human written training set containing a lot of em dashes.

Thus, all machine learning based approaches are pretty doomed to fail.

checkifmytextisai.com is far superior, since it is purely based on vibes and mimicking intuition instead of trying to make something that's fundamentally impossible. Now, I'm not saying that it'll always be right, but start trying it out on articles, tweets, conversations, etc. I gurantee you that you will find a better result from it than any machine learning based 'AI TEXT DETECTION TOOL' that currently exists on the market.

Who built this

I’ve spent thousands of hours working directly with large language models. Not casually. Professionally, obsessively, across every domain, tone, and style you can think of. Coding, creative writing, marketing copy, academic papers, casual conversation. That volume of hands-on time builds an intuition that most people simply don’t have yet.

After enough exposure, you stop needing to look for em dashes or the “It’s not X. It’s Y” pattern. You start recognizing deeper things: how AI distributes emphasis too evenly, how it sounds thorough while never committing to anything specific, how it hedges in ways no real person would. I distilled all of that intuition into the prompt engineering that powers this tool.

If you would like to learn more about my specific background, check out my pinned article on Twitter/ X!

Yes, this is actually free

No signups. No paywalls. Every analysis runs a full LLM call, which costs real money per request. There’s a lot of room for this approach to get even more accurate that I'm not building in right now, since everything is on my own dime and I want it to be publically accessible.

Even so, it already outperforms traditional ML-based detectors by a significant margin.