Diagnosis · LinkedIn post
Why your LinkedIn post sounds like ChatGPT
The instinct is to blame the model for inventing this voice. LinkedIn had already invented it.
LinkedIn posts read as ChatGPT-written because the causation runs backwards from what people assume: the hook line, bold-labeled bullets, and closing engagement question were LinkedIn's own creator-economy convention years before any general-purpose model existed. Ghostwriting courses taught that shape, creators copied it because they believed it worked, and the resulting posts became training data. The fix is breaking the shape itself, not swapping the words inside it.
Why this happens to LinkedIn post in particular
By the early 2020s, a whole cottage industry of creator coaching had built a template around driving comments on the feed: a one-line hook with a hard stop, a white-space gap, three or four bold-labeled bullets for skimmability, and a closing question that begs an easy reply. None of that required a language model; it was taught in paid courses and copied post to post because creators believed it worked. When general-purpose models trained on the open web, LinkedIn's public posts and the how-to-write-them advice around them were part of the corpus, so a model absorbed the format as what a LinkedIn post looks like. It is not improvising a chatbot voice. It is quoting a template LinkedIn's creator economy built first.
The tells that show up here
Not every AI tell appears everywhere. These are the ones that cluster on this surface, and what it is about the format that produces them.
- not-x-its-y
- The reframe that turns a professional stumble into a lesson learned was a growth-coaching formula taught in posts and courses for years before a model existed to imitate it; the model learned the coaches' trick, not an original one.
- bold-header-list
- The bold one-word label followed by a bullet was recommended in paid LinkedIn-growth courses specifically as a skimmability tactic, so the model reproduces a particular creator-economy convention, not a generic bureaucratic habit.
- chatbot-artifact
- A different failure from the engagement question above: sometimes the chat interface leaks through literally, a draft closing on a stock offer to keep assisting or a reflex reassurance that the reader can always ask for more, phrasing lifted straight from an assistant register rather than reworded into LinkedIn's own comment-bait convention. That is not the model imitating LinkedIn's format. It is the model forgetting to leave the chat voice behind.
- corporate-jargon
- Phrases like double down, move the needle, and circle back cluster in LinkedIn posts because the platform's own professional register already ran on this vocabulary before a model existed to reproduce it; the checklist tracks it but does not score it, since human LinkedIn writers use the same words unprompted.
Before and after
A generic AI draft on the left. On the right, the real output of running that draft through underslop, checked so it adds no name or number the draft did not already carry. The voice is The Dry Contrarian, unimpressed with the performance
Vulnerability is a strength. 🙌 Last month I almost missed a deadline because I was too proud to ask for help. It's not a failure, it's a wake-up call. Here's what I learned: - **Humility:** Asking for help is not weakness. - **Trust:** Your team wants to see you succeed. - **Growth:** Every stumble is a setup for a comeback. What's a moment vulnerability taught you something? Drop it below! 👇
Vulnerability is a strength. Last month I nearly blew a deadline because I couldn't bring myself to ask for help. Not a failure, exactly—a wake-up call. What I took from it: humility means admitting you don't have it all figured out; trust means your team actually wants you to do well; growth means the stumble is part of the arc, not the end of it. I still don't know that I'd call the feeling strength. But it was something.
How to fix it yourself
- Delete the hook line and the white-space gap Cut any opening line built purely to stop a scroll, especially one that ends on a period with nothing else in the sentence. Start with the actual event.
- Merge the bold-labeled bullets back into the story If each bullet restates a piece of the same event, write it as continuous sentences instead. Keep a real list only for genuinely separate, parallel items.
- Cut the reframe if the plain version already lands Check whether the post reframes a plain fact as something grander through a paired negation-and-correction sentence. If the plain fact already carries the same weight, the reframe was decoration, not information, and it goes.
- Replace the engagement question with an actual stopping point End on the last true detail instead of a question aimed at the comment count. If you want a real conversation, ask a specific question about the reader's situation, not a generic prompt any post could carry.
Questions
Why does ChatGPT sound like LinkedIn?
The causation runs the other way: LinkedIn's hook-bullets-question format was a creator-economy convention years before general-purpose models existed, taught in paid courses and copied widely because creators believed it worked. A model trained on the resulting posts learns the format as the default answer to a request for a LinkedIn post.
How to spot a ChatGPT LinkedIn post?
Look for the shape more than any word: a one-line hook standing alone, a white-space gap, three or four bold-labeled bullets, and a closing question aimed at the comment count. A post missing all four of those rarely reads as machine-made, whatever words fill it.
Why do LinkedIn posts sound like that?
Creator-coaching courses spent years teaching a template built to drive comments: the hook, the bullets, the question. Models absorbed that template from the posts and the how-to guides around it, not the other way around.
How to make AI-generated content sound more human?
On LinkedIn specifically, breaking the hook-bullets-question shape matters more than changing any word inside it. Merge the bullets back into sentences, cut the engagement question, and let the actual event carry the post instead of the scaffolding built to trigger comments.
Is the hook-bullets-question format always a problem?
No, and it worked before AI for the same reason it still works: it is easy to skim. The tell is using it on a post that has one real idea, where the format adds performance the idea never asked for.
This page explains the cause. underslop is the edit: paste up to 500 words on the free tier, and it marks the tells it could not remove. Your text is held for 0 seconds.