AI copywriting can be good. What gives the bad kind away, even after you fix the words, is the shape.
If you use AI to write your company's posts or pages, you've probably wondered whether people can tell. Some of them can. "People can tell. Not everyone can tell, but some people can," is how one person put it in Reddit's digital marketing forum this year.
I don't have as big a problem with AI copywriting as a lot of people do. I think it can be done pretty well, and there's a time and a place for it. But I can usually spot the bad kind, and what gives it away for me is the shape and the tone: a line break after every sentence, a lot of emojis, a post that runs incredibly long, and a casual voice that sounds like somebody told it to be casual.
Megan McTaggart17 November 2026

The people most likely to notice use AI themselves
In older studies (2021, 2023), casual readers did about as well as a coin flip at telling AI writing from human writing. But in a 2025 study, people who often use ChatGPT for writing work caught AI-written articles 92.7% of the time, and rarely flagged a human one by mistake (Russell, Karpinska and Iyyer, 2025). It was a small panel, five readers on a first batch of 60 short articles, so the direction matters more than the exact number.
My bet is a lot of owners and marketing leads are in that group by now. They know what the default output looks like because they've deleted a lot of it.
Swapping the words doesn't hide it
Vocabulary was the most common clue the people in that study gave, the words a person rarely types. That part is real, and it's also the easiest part to change. Here's a made-up post, the kind I mean:
Most companies get marketing wrong.
Here's why.
They focus on tactics. 🚀
Not strategy.
The result? Wasted budget.
Every sentence gets its own line, and it asks a question so it can answer it. You could swap every word and still know where it came from.
The tone gives it away too. Another made-up one:
Okay, real talk. I used to think brand voice was fluff, lol. Turns out? It's everything.
You can tell somebody typed "make this sound conversational." It feels forced, and the people who write with these tools every day tend to hear it.
I don't have as big a problem with AI copywriting as a lot of people do.
Why it keeps happening
My read is that if you start with a topic and nothing else, you get back something close to what everyone else got. In one study, writers given ideas by an AI wrote stories that were judged a little better, and the stories as a group came out more alike than the ones written without it (Doshi and Hauser, Science Advances, 2024). A different tool won't change that much. In one 2025 study, different models gave strikingly similar answers to the same open-ended questions.
Asking the tool to sound more human changes the tone and leaves the point where it was, which is usually the same point everyone else is making.
How to use it so it doesn't read like that
Start with what you actually think, and one real detail, like something a customer said or a number you have. Then let the tool help with research and with putting it in order. Google says AI is useful for researching a topic and adding structure, and that anything it helps write should be checked for accuracy before it goes up (Google Search Central). Appropriate use of AI isn't against its search guidelines, and that's been Google's line since 2023.
Then edit the shape as well as the words. Put the sentences back into paragraphs and cut what it doesn't need. Take out anything that answers its own question. Then read it out loud and change anything you wouldn't actually say.

