On 30 July 2026, LinkedIn added a new option to the three-dot menu on every post: "Seems like AI slop." Tap it and the post vanishes from your feed. LinkedIn also records that a real person looked at that writing and decided it was not worth their time.
Most of the coverage read this as LinkedIn cracking down on AI. It is not that. The detail that matters got buried under the headlines, and it changes what you should actually do about your own posts.
What the LinkedIn AI slop button actually does
Three things happen when a reader taps it. Only one of them is about you, and it is the one nobody mentions.
- The post disappears for that reader. Immediate and local. Roughly what "not interested" already did.
- The report trains LinkedIn's classifier. Chief product officer Hari Srinivasan has said the reports are used to tune the models that identify low-quality AI content.
- Your reach outside your network drops. Reported posts get pulled back in feed recommendations. That is the same mechanism that already sat behind "not interested," which is why the effect is real but rarely dramatic from one tap.
LinkedIn is also testing a private flag inside your own analytics. It tells you that readers are finding your content inauthentic. It is not a strike, nobody else can see it, and it does not appear on your profile. Think of it as someone quietly telling you that you have started repeating yourself.
Worth saying plainly: one report will not sink a post. The signal is meant to work in aggregate, across many readers and many posts. What should worry you is a pattern, not an incident.
It is not an AI detector
Nothing is scanning your post to work out which tool wrote it. LinkedIn's chief product officer put it plainly: "AI and slop are not the same thing." LinkedIn does not penalise people for using AI, and the authenticity flag does not run on an automated detector. It runs on what readers report.
That difference is the whole thing. A detector asks whether a machine wrote it. A reader asks whether it was worth their time. Only the second question has a button attached to it.
It also explains a decision LinkedIn made in the same week, which got almost no coverage. They retired "Enhance your post", the feature that rewrote your draft for you, and replaced it with a proofreading tool that leaves your voice alone. A company at war with AI writing does not ship a better writing assistant. A company worried about everyone sounding identical does exactly that.
How much of LinkedIn is actually AI-written
The detection firm Pangram estimated in 2026 that around 41% of long-form LinkedIn posts and 30% of short-form posts were likely machine-drafted. It also found that LinkedIn accounted for roughly 62% of all the AI content it scanned across major social platforms.
Treat those figures as directional. They come from an AI detector, and detectors are unreliable enough that we would not build a product on one. But look at what LinkedIn did with that reality. They did not ban it. If four in ten long-form posts were AI-assisted and LinkedIn believed that was the problem, the button would say "report AI." It says "seems like AI slop."
What gets flagged as AI slop
Readers react to four things before they get to your vocabulary: whether there is anything real in the post, what shape it is, whether you take a position, and whether the body earns the first line. A banned-word list reaches none of them.
Here is what each one looks like on the page.
| What readers react to | What it looks like |
|---|---|
| Nothing checkable | No number, no name, no date, no moment that actually happened. Advice that would be true for anyone in any industry. |
| Template shape | A short standalone opening line, a blank line, one-sentence paragraphs, exactly three of everything, a tidy closing aphorism, "Agree?" |
| No position | Balanced coverage nobody could disagree with. No cost admitted, nothing the author could turn out to be wrong about. |
| No payoff | The first line contains the whole point. The rest restates it in fresh words. |
Word choice comes after all four. You can strip every instance of "leverage" and "unlock" out of a post and still have something a reader scrolls past, because the reason they scrolled past was that the post did not say anything.
Why buzzword checkers miss it
Almost every anti-slop tool on the market reads word choice and stops there. Banned-word lists, em dash strippers, a counter for hashtags and hype adjectives. Those are all worth having. A post can also pass every one of them and still be exactly the thing the button was built for.
The reason is simple. Word choice is the one instruction a language model follows perfectly on the first ask. Tell it to avoid "delve" and it will never write "delve" again. The template underneath survives untouched, and the template is what a reader recognises before they have read a word of it.
This is also why running a flagged post back through a rewrite rarely helps. A second pass through the same kind of model returns the same silhouette in fresh vocabulary. What actually changes a post is adding the thing only you could have written, then breaking the shape it arrived in.
What to do instead
None of this requires you to stop using AI. LinkedIn has said outright that it does not penalise you for it. Four habits do most of the work.
- Bring one particular the model could not have. A figure from your own week. A client's name. The date something broke. A model cannot invent the detail that makes a post yours, and a post without one reads as generic no matter which model wrote it.
- Let the structure follow what you have to say. If your opening line sits alone above a blank line, you have already told the reader which template you used before they reached your second sentence.
- Take a position with a cost attached. Something you could turn out to be wrong about. Balanced coverage is what nobody argues with and nobody remembers.
- Cut anything the reader could have predicted. If your first line contains the whole point, there is no post underneath it. Move the surprising part up and delete what was restating it.
A useful test before you publish: could a competitor post this exact text under their own name and have it fit? If the answer is yes, you have written the industry's post rather than yours.
Check a post before you publish it
We built a free AI slop checker that reads a post the way a reader would. It scores the four things above plus the surface tells, quotes the worst line back to you with the reaction a reader would have, and asks the one question that would make the post specific. There is no login. Paste anything, including drafts you wrote yourself or in ChatGPT.
Two things it deliberately will not do. It will not tell you whether AI wrote something, because detectors are unreliable and biased against people writing in a second language. One Stanford study found 61% of TOEFL essays by non-native English speakers were wrongly flagged as machine-written. And it will not promise a post is safe from the button, because the button is pressed by people and people disagree with each other.
What it gives you is the line a reader is most likely to bounce off, and what to do about that line.
For the background, start with what AI slop is and whether your own post is likely to get flagged. If you want to see how FeedBoss handles all of this inside its own drafts, we wrote that up in how FeedBoss prevents AI slop.
FAQs
What is the "Seems like AI slop" button on LinkedIn?
A reporting option added on 30 July 2026, in the three-dot menu on any post. It hides the post from your feed, feeds LinkedIn's classifier training, and reduces how far the post travels outside the author's network.
Will LinkedIn penalise me for using AI to write posts?
No. LinkedIn's chief product officer has stated that "AI and slop are not the same thing" and that the platform does not penalise AI use. What gets reduced reach is content people report as low-value, which is a judgment about the writing rather than about the tool that produced it.
Does LinkedIn use an AI detector to flag my posts?
Not for the authenticity flag. LinkedIn has said that flag runs on signals from members rather than an automated detector, on the grounds that plenty of people use AI to refine their thinking without producing slop.
How do I stop my posts being reported as AI slop?
Put something checkable in every post: a number, a name, a date, a specific moment. Let the structure follow the content instead of the hook-and-blank-line template. Take a position that has a cost. Cut anything the reader could have guessed from your first line.