You’ve felt it. You’re scrolling, you land on a post, and before you’ve finished the first sentence some part of your brain has already filed it under “a robot made this.” You don’t even decide to think it. It just happens.
That reaction is worth taking seriously, because your readers have it too. And once they’ve had it about your writing, the rest of the post barely matters.
So what are they actually picking up on?
The vocabulary that nobody says out loud
There’s a small set of words that real people rarely reach for but language models love. Delve. Leverage. Robust. Seamless. Underscore. Tapestry. On their own they’re harmless. Stacked three to a paragraph, they read like a press release.
The deeper tell isn’t any single word. It’s that the writing sounds like it’s describing something rather than saying it. “This approach can significantly enhance your productivity” is a sentence with no person behind it. “I got an hour of my morning back” is a person.
The rhythm is too even
Machine-written text tends to march. Every sentence is roughly the same length. Every paragraph is three or four lines. Every list has exactly three items, balanced like a tripod.
Human writing is lumpy. A long, winding thought, and then a short one. A fragment, even. We speed up when we’re excited and slow down when something matters. That unevenness is information. When it’s sanded flat, the reader feels the flatness even if they can’t name it.
It’s confident about nothing in particular
The most common giveaway is that the post says a lot of agreeable things and risks nothing. It “explores,” it “highlights the importance of,” it reminds you that consistency is key. None of it is wrong. None of it could only have come from one specific human who has actually done the thing.
Real writing has edges. It says this worked and that was a waste of money. It admits the author was wrong about something for two years. A model averaging across the whole internet has no edges by default, because the edges are exactly what got averaged out.
Why the slop quietly costs you
Here’s the part people underestimate. The damage isn’t that a generic post is boring. It’s that it teaches the reader something about you: that you’ll hand them something you didn’t really make.
Attention is a trust transaction. Someone gave you thirty seconds. If they spend it on text that any of ten thousand accounts could have posted, they learn to skip you next time. Not angrily. Just automatically, the same way you scrolled past the last one.
You don’t get flagged as a bad writer. You get filed as noise. And noise is much harder to climb out of than mediocrity, because nobody’s even reading closely enough to notice you got better.
The fix isn’t writing less with AI
The reflex is to swing the other way and write everything by hand again. That’s not the lesson. The lesson is that the raw material has to be yours.
A model is genuinely good at structure, tightening, and getting a draft unstuck. It is bad at having lived your life. When the input is a real opinion, a real story, a thing you actually noticed this week, the output stops sounding like a machine, because the substance underneath it isn’t generic. The words can be polished. The point of view can’t be borrowed.
That’s the whole game. Sound like a person by starting from something only you could have said.
Talk through your ideas and publish posts that sound like you.
Claim a founding spot