You have almost certainly scrolled past a post thinking "that's AI" without being able to say exactly why. That reflex is now widespread on LinkedIn, and it is expensive: once a reader has attached that label to your text, they stop reading and start checking.
The good news is that what gives a generated post away is not mysterious. It is always the same handful of markers.
What gives an AI-written post away
Flawless transitions. "But that's not all." "And here's where it gets interesting." Nobody writes like that on their phone between two meetings. Those hinges are the signature of a text that was assembled rather than thought through.
No verifiable detail. A human post says "on Tuesday a client asked me why their invoice had doubled". A generated post says "recently, a client raised a concern". The second is smoother, and perfectly interchangeable with ten thousand others.
Symmetry. Three paragraphs of identical length, three numbered points, three examples. Real writing is uneven: you linger on what matters and rush the rest.
A disembodied ending. "In short, consistency is key." A conclusion you could paste onto any other post is not a conclusion.
The real problem is the starting point
The methodological mistake is asking AI to produce the finished post. From an empty prompt, a model can only produce the average of what it has read — which is precisely the style your readers have learned to spot.
An AI has no idea what conversation you had yesterday, what sentence your client used, or what annoyed you this week. That is exactly where good posts come from.
So flip the order: bring the raw material, let the AI handle the shaping.
The method, in four steps
1. Note the raw idea, without polishing. Three lines is enough: what happened, what you took from it, why it might help someone else. That is your content, and it is already unique.
2. Give your style as examples, not adjectives. Asking for "professional but approachable" means nothing to a model. Feeding it five of your past posts does. That is the principle behind style analysis in Ego: your previous publications act as the reference, and your sentence length, punctuation and way of opening a text are extracted from your own writing.
3. Treat the output as a draft. The first pass exists to give you a structure, not a publishable text. Cut the grand transitions, replace a generality with your specific example, break the symmetry.
4. Add the sentence the AI could not have written. A number from your own business, the software you actually use, the exact objection someone raised. That is what makes a post impossible to reproduce.
A simple test before publishing
Reread your post and ask: could a competitor publish this exact text with only their name changed?
If the answer is yes, it is not a post, it is a press release. Add what belongs only to you — a date, an amount, a mistake you made — and the problem disappears.
What AI genuinely does well
This is not an argument against the tool. Used in the right place, AI saves a great deal of time:
- rewriting a hook ten times to find the angle that lands;
- cutting 2,000 characters down to 1,200 without losing the argument;
- noticing that your second paragraph repeats the first;
- turning one idea into three tonal variants so you can pick the one that sounds like you.
Those are editing tasks, and AI is excellent at them. Coming up with the idea stays on your side — which is good news, because it is the only part nobody can copy.
The takeaway
The problem was never AI itself, but the order of operations. A post written from nothing will always ring hollow. A post written from your material, shaped by AI and then reworked by you, simply reads like a well-written text — because it is one.
Generate your LinkedIn posts with Ego
Try for free
