Why your LinkedIn posts sound like AI (even when you wrote them)
The tell isn't the words. It's that the post could have been written about anyone. Here's the actual mechanism, and how to break it.
There is a particular flavour of LinkedIn post that makes people scroll past on reflex. Short punchy lines. A hook that promises a lesson. A three-beat story with a tidy moral. Something like: I fired our biggest client. Everyone said I was insane. Here's what happened next.
The reflex reading is that it sounds like AI wrote it. But plenty of those posts are written by hand, by real founders, at eleven at night. So the tell is not the words. It is something more specific.
The actual tell: substitutability
A post reads as generic when you could swap the author's name for any other founder in the same category and lose nothing. That is it. That is the whole mechanism.
"Consistency beats intensity" is true and it is also substitutable. Twelve thousand people could have posted it. "We spent a quarter's ad budget on Instagram for a B2B client and got zero pipeline, because B2B buyers don't make vendor decisions on Instagram" is not substitutable. It happened to one person.
The question is not "is this well written". It is "could anyone else have written this". If yes, it will read as generic no matter how good the prose is.
Why templates guarantee it
Every LinkedIn tool that starts from a template is optimising for the substitutable case by construction. A template is, definitionally, the part of a post that is true for everyone. Fill in the blanks and you get the average of every post that ever fit the shape.
This is why the output does not improve when you write a better prompt. You are not fighting the model's writing ability. You are fighting the fact that you gave it nothing that only you know.
The specificity ladder
When a draft feels flat, it is almost always sitting too high on this ladder. Walk it down:
A claim: "Founder-led sales works."
A claim with a category: "Founder-led sales works better than hiring reps too early."
A claim with a number: "We tried to hire two reps at 400k ARR and it set us back six months."
A claim with a scene: "Our second rep quit on a Tuesday. He said he could not sell something he had never seen a customer use. He was right, and I had no answer."
Rung four is not better writing. It is the same event with the details left in. Most founders delete exactly those details in the name of professionalism, and delete the post's reason to exist along with them.
The awkward part
The specific version costs something. Naming the quarter you burned means admitting you burned it. That discomfort is a reliable signal: the posts that feel slightly risky are usually the ones worth publishing, and the ones that feel completely safe are the ones nobody will finish.
You do not need better prose. You need to stop sanding the edges off the things that actually happened.