How to tell if a LinkedIn post was written by AI, before a tool does it for you

You've probably had this moment: you're halfway through a LinkedIn post, nodding along, and somewhere around the third tidy paragraph a small voice says "wait, did a person actually write this?" You can't always prove it. But you can usually tell, once you know what to look for. I spend a lot of time reading these things, because detecting them is literally what clutters a LinkedIn feed in the first place, so I've ended up with a mental checklist. Here it is, in plain language, no tool required.

What are the clearest tells in an AI-written LinkedIn post?

Four things show up together more often than chance would explain: a hook built on a reversal ("I used to think X. Then Y happened."), a structure that reads like a listicle even without bullet points, a sentence rhythm that never breaks stride, and a closing question aimed at nobody in particular. Any one alone proves nothing. Three or four together is a strong signal.

The reversal hook

"I used to believe hard work was the only thing that mattered. Then I watched a junior hire outperform me in a week, and everything changed." It's a clean little story shape, which is exactly the problem — real anecdotes are usually messier than that. A specific Tuesday, a named project that fell apart for a boring reason, a detail that doesn't serve the lesson. The reversal hook is the lesson wearing a story costume.

The invisible bullet points

Three short paragraphs, each starting with a bolded phrase, each exactly two sentences, building to a one-line takeaway. Nobody drafts that rhythm by accident. People write in bursts — a long thought, then a short one, then they trail off and come back to it. A post with no uneven paragraphs is a post that was assembled, not drafted.

The sentence that never stumbles

Every sentence lands the same length, the same confident cadence, no fragments, no "anyway," no parenthetical aside. Actual writing wobbles. It has a sentence that's too long because the person was still figuring out what they meant while they typed it. Smoothness at that level is the tell, not the sentences themselves.

The question nobody asked

"What's been your experience with this? Let me know in the comments 👇" tacked onto the end of a post that already made its point. It's an engagement prompt, not a real question — the post isn't actually curious about your answer. A real person asking usually asks something narrower, like "has anyone else dealt with this exact thing," not a generic invite to the whole feed.

Does spotting these tells mean the post is bad?

No. None of this is a quality judgment — the actual problem is volume. When enough posts share these patterns, they start to look alike, and a feed full of lookalikes buries the one post from someone you actually follow who had something real to say. The tells matter because they show where your attention is going, not because the writing underneath is worthless.

What do you do once you've spotted a pattern like this a few times in a row?

Noticing is already useful on its own — you scroll past faster, and you stop mistaking volume for relevance. If you want it automatic, Zenfeed's Rest from AI does the same kind of pattern-matching on a timer, so the rest of your feed gets a chance to surface while it's running. It's not a verdict on any one post, which is also why it's a timer and not a permanent block.

It works by scanning each post's text (and image alt text, so nothing hiding in a graphic slips through) against a list of AI-related phrases and words, not by reading the prose style. So it won't catch every post you'd personally flag by eye using the tells above, and it isn't trying to — it's a volume valve, not a style critic. Here's what's actually behind it on LinkedIn and X:

None of this requires reading every post with a magnifying glass. Most days, the manual checklist and the automatic timer end up doing the same job at different speeds — one in your head, one in the background.