Feed behaviour
How to spot AI-generated posts on your work feed
- From:
- Hesham Barr, founder
- Published:
- Updated:
The short answer: look for confident words with no specifics. AI-written posts tend to have a punchy one-line opening, short dramatic paragraphs, tidy lists of three, and a closing "What do you think?", but the real giveaway is that nothing in them could only have come from the person who posted it. No numbers from their work, no named tools, no mistakes they made.
How much of what you read is AI-written?
More than you might think. Originality.ai, which sells AI detection, studied longer public LinkedIn posts from 2018 to late 2024 and classified about 54% of recent ones as likely AI-written. Its July 2026 sample of 5,000 posts of 100 words or more put the figure at 81%. The company also found posts got much longer after ChatGPT launched at the end of 2022.
Two caveats are fair. AI detectors are not perfect, and the study looks at a slice of public posts, not every feed. But nobody who uses these networks is surprised by the trend.
What are the signs a post was written by AI?
None of these proves anything alone. Several together, with no real detail, is a strong hint.
The shape
- A one-line hook: "I almost didn't post this."
- Single-sentence paragraphs, each on its own line, building to a "lesson".
- Lists of exactly three: "Clarity. Consistency. Courage."
- A neat closing question: "What would you add?"
The phrasing
- "Here's the thing."
- "Let that sink in."
- "This is not X. It's Y."
- "In today's fast-paced world."
- Lots of words like "unlock", "leverage", "navigate", "journey" and "landscape".
The substance, or lack of it
- No specifics. No figures, dates, tools, places or names. "We transformed our process" rather than "we moved invoices from email to the finance portal and cut late payments by a third".
- No friction. Real work stories include things that went wrong. Generated ones resolve neatly.
- Interchangeable. If you could swap the author's name for anyone in a similar job and it would still read the same, it probably was not drawn from their experience.
The comments
AI is now writing replies too. Watch for comments that restate the post in different words ("Such a powerful reminder that leadership is about people!") with nothing added.
Is it wrong to use AI to write posts?
Not as such. Many people use AI to tidy grammar, to write in a second language, or to turn rough notes into something readable. That is fine.
The problem is posts with nothing behind them. A network for work is valuable because people share what they actually know. Text that only sounds like knowledge wastes everyone's time, and it drowns out the posts that do have something to say.
Why does it matter more on a work network?
On a general social app, a hollow post is a minor annoyance. On a work network you are using what you read to make decisions: which tool to try, how to handle a problem, who to trust. If you cannot tell real experience from generated filler, the network stops being useful.
What can a network actually do about it?
Platforms have started to respond. LinkedIn says it reduces the reach of generic AI-looking content and has been rolling out a "seems like AI slop" option so members can flag it.
At Office Worker we look at it from the other end. Rather than trying to prove who wrote a post, our checks ask whether it is about work and whether it says something. Posts that are all hook and no substance, or that exist mainly to collect reactions, are caught before they reach a feed. And because the feed is not a reaction contest, there is little reason to churn out filler in the first place.
You can read more in what a work-only network looks like and what is engagement bait.
How do you write a post that clearly isn't AI?
Simple: put in things only you could know.
- A number from your work. "Our returns dropped from 9% to 6%."
- A specific tool, place or constraint. "On a 40-person team with no data engineer."
- Something that went wrong. "The first version made it worse."
- A plain opening. Say the point in the first sentence.
- No closing hook. If you want answers, ask a real question.
Frequently asked questions
- How much of LinkedIn is written by AI?
- Estimates vary, but one detection company, Originality.ai, classified 54% of longer English LinkedIn posts as likely AI in late 2024, and 81% of a 5,000-post sample in July 2026. Detectors are imperfect, so treat these as rough indicators of a clear trend.
- How can you tell if a post was written by AI?
- Common signs include a one-line hook followed by single-sentence paragraphs, lists of three, phrases like 'here's the thing' and 'let that sink in', confident claims with no specifics, and a closing question such as 'What do you think?'. No single sign proves anything; look for several together and, above all, for a lack of real detail.
- Is using AI to write a post wrong?
- Not in itself. Many people use AI to fix grammar or to write in a second language. The problem is posts with nothing behind them: generated text that sounds confident but shares no real experience, numbers or examples.
- What is LinkedIn doing about AI slop?
- LinkedIn says it reduces the distribution of generic, repetitive content that appears AI-generated, and has been rolling out a 'seems like AI slop' feedback option for posts and ads.