I’ll Instantly Know A Writer Used ChatGPT When I See This

I’ve started noticing the same commenter joining the discussion in hundreds of articles. Since he uses a pseudonym, and I don’t know his actual name, I’ve started referring to him as “AI Detective Dylan” in my head. Every few days, I’ll finish reading an article, scroll down to the comments, and find Dylan explaining why the piece was “obviously written by ChatGPT.”

Whenever I see one of Dylan’s comments, the pattern is usually the same. He regularly breaks down articles sentence by sentence, pointing to phrases, transitions, and word choices he believes reveal AI involvement.

“Once you’ve seen enough AI writing, it becomes easy to spot,” Dylan wrote. “Anyone who disagrees with me is delusional!”

Last month, I watched Dylan accuse one writer of using ChatGPT because the article contained “a few metaphors.” A few days later, I saw he made the same argument under a completely different article written by someone else. Meanwhile, he seemingly ignored writers publishing 100,000 words a day on multiple platforms while somehow keeping up with social media, client work, and full-time jobs.

After a while, I started noticing that some writers seemed to be publishing way faster than I could read. One article in the morning. Three more by lunch. And for good measure, about 300 LinkedIn posts in between. Then, I’d open my inbox the next day and discover they’d done it all over again.

At first, I assumed some people were simply more productive than me.

But then I started paying attention.

A few months ago, I clicked through one creator’s archive and kept scrolling. By the time I reached the previous week, the word count was already somewhere near seven figures. Another creator had somehow published even more. Around the same time, I kept seeing people argue about metaphors, transition phrases, sentence structure, and AI detector scores. The debates never seemed to lead anywhere. The publishing schedules, on the other hand, were sitting right there in public.

I don’t know about you, but I can’t operate in that universe. Some days, I sit down and knock out a draft of 1,500 words before lunch. That feels great. But other days, I spend fifty minutes staring at a paragraph, decide the third sentence is the problem, rewrite it, decide the fourth sentence is now the problem, and somehow end up back where I started. Heck, on a bad day, a determined tortoise could probably complete a respectable lap of my home office before I finished rewriting the opening sentence.

Here’s something people like AI Detective Dylan don’t want to admit…

AI detectors aren’t reliable indicators of AI, and neither are many of the supposed “writing tells” people keep pointing to. Meanwhile, some creators are publishing 100,000 words a day while also keeping the rest of their lives intact. So, at a certain point, basic arithmetic is a lot more useful than punctuation analysis.

When people discuss how AI is changing writing, the conversation usually comes back to detection, disclosure, and authenticity. The assumption is that better software will eventually make it easy to separate human work from AI-assisted work. Looking at the current tools, though, that seems far from guaranteed.

As things stand, AI detectors regularly disagree with each other. I’ve watched writers paste the same article into multiple detectors and receive completely different results. So out of curiosity, I pasted the Declaration of Independence into two separate AI detectors.

Guess what?

The scores came back wildly different.

ZeroGPT estimated 92.5% AI. And GPTZero gave the exact same text a score of 2% AI. So either the Founding Fathers were time travelers with access to large language models, which would certainly please my inner conspiracy theorist. Or, we can throw our hands up and admit these tools aren’t exactly foolproof.

For what it’s worth, writing pays a decent chunk of my bills these days. I run a publication called The Daily Draft, have more than 150 paid subscribers on Substack, and have been fortunate enough to see my work appear in places like Business Insider. Besides Grammarly‘s spellchecker catching the occasional typo, and occasionally using speech-to-text dictation software if I want to rant aloud, I don’t use AI assistance in my writing process.

Author and entrepreneur Seth Godin has written extensively about the connection between writing and thinking. Looking at most writers I know, that relationship is hard to miss. The words on the page are only one part of the process. Reading, researching, outlining, editing, and changing your mind all count too.

The bigger question is what happens next…

Personally, I’d welcome some sort of human-created stamp on books and other published work. Something simple but universal. The literary equivalent of a copyright symbol. But until then, readers are going to have to rely on observation, context, and common sense. Because if the Declaration of Independence can confuse an AI detector, blind faith in software definitely isn’t the answer.

Did you find this article interesting? Consider reposting this story so more people can read it.