On 30 July 2026 LinkedIn added a “Seems like AI slop” option to the three dot menu of posts and comments. A flagged post or comment disappears from the feed of the person who reported it, and the report trains LinkedIn's classifiers for low quality, AI written content. Chief product officer Hari Srinivasan said LinkedIn already blocks hundreds of thousands of automated slop comment attempts every day. Using AI isn't banned. Sounding generic is what gets you flagged.
You've seen the comments. “Great insights, thanks for sharing!” under a post about a funeral. A five line paragraph that restates the post in fancier words, then asks “What are your thoughts?” Most people scroll past. Since this summer, they can also report it.
We build an AI commenting tool, so this change landed on our desk the day it shipped. Full disclosure, then: we have skin in this game. That's also why we read every announcement, help page and press quote we could find, and wrote down only what LinkedIn actually said. Where something isn't documented, we say so.
What is LinkedIn's “Seems like AI slop” button?
It's a report option. Open the three dots at the top right of any post or comment and you'll find “Seems like AI slop” next to the usual choices. LinkedIn announced it on Thursday, 30 July 2026, in a post by Hari Srinivasan, and the story ran the same day in TechCrunch and Social Media Today, then in Fortune and The Register.
Is it live for everyone? Not quite. Srinivasan described it as something LinkedIn is “ramping”, so some accounts saw it weeks before others. A LinkedIn spokesperson told reporters it covers all posts and comments, which matters a lot if most of your LinkedIn activity is commenting rather than posting.
Why now? LinkedIn has a slop problem and it knows it. An analysis by the AI detection company Pangram, quoted widely in the coverage, found that roughly 30% of short posts and over 40% of long posts on LinkedIn read as fully AI generated. Those are Pangram's numbers, not LinkedIn's. Still, nobody at LinkedIn argued with them in public.
What happens when someone flags your post or comment?
Less than people fear, and more than nothing. Here's the honest version.
The person who flags it stops seeing it. The post or comment drops out of their feed, and LinkedIn thanks them for the feedback. That's the only effect LinkedIn has described for a single report.
The report becomes training data. Srinivasan was clear that slop is hard to define and the definition keeps moving, so the reports are there to tune LinkedIn's models. Those models already decide what gets shown to people outside your network, which is exactly the audience a good comment is supposed to reach.
You, the author, don't get a public mark. Nothing appears on your profile and nobody else sees a label. LinkedIn did say it will test a private flag in your analytics that tells you when members feel a post seems inauthentic or heavily AI generated. Think of it as a smoke alarm only you can hear.
Does one flag hurt your reach? LinkedIn hasn't said. There's no published threshold, no documented penalty, and anyone who quotes you a number (“three flags and you're shadowbanned”) is guessing. What we can say is that the reports feed the same systems that rank comments and suggested posts, so a pattern of flags is unlikely to help you.
Everything LinkedIn announced alongside the button
The slop button was one of several changes that shipped together. They make more sense as a set.
| Change (July 2026) | What it does | Who notices |
|---|---|---|
| “Seems like AI slop” report | Hides the post or comment for the reporter and sends a training signal to LinkedIn's models | The person reporting; LinkedIn's ranking systems |
| New slop classifiers | Filter low quality and AI slop out of suggested posts from outside your network | Everyone's feed, quietly |
| Private “seems inauthentic” flag | A test flag in post analytics when members feel your content is heavily AI generated | Only the author |
| “Enhance your post” removed | LinkedIn's own AI rewriter goes away, replaced by a proofreading tool that, per Srinivasan, doesn't change your voice | Anyone who used the rewriter |
| Automated comment blocking | LinkedIn says it blocks hundreds of thousands of automated slop comment attempts a day, plus billions of other automation attempts over a couple of months | Comment bots, mostly |
| Verification push | More profile and page verification; over 100 million members have verified their identity | Everyone, via the badge |
| Block comments from company pages | Lets you hide comments from pages you no longer want to hear from | People tired of brand replies |
Notice the direction. LinkedIn removed its own AI rewriter and replaced it with a tool that keeps your voice. That's the clearest statement of policy in the whole announcement, and it's written in product, not in a press release.
Fezerli reads the whole post, images and PDFs included, writes from your own past comments and rewrites any draft that slips into the usual AI tells before you see it. Nothing goes out until you press Post.
Is using AI to write LinkedIn comments against the rules now?
No. Srinivasan said plainly that people who use AI to refine their thoughts should keep doing it. The button targets output that reads like nobody was home, whatever produced it.
Automation is a different question, and it's worth keeping the two apart. LinkedIn's help page on prohibited software points to section 8.2 of the User Agreement, which forbids bots and unauthorized automated methods that comment on, like or share posts, or otherwise drive inauthentic engagement. So:
- You read a post, an AI tool drafts a reply, you edit it and press Post. That's you using a writing aid, the same as a spellchecker with opinions.
- A script reads your feed and posts comments while you sleep. That's automation, and LinkedIn's rules aren't fond of it whether the words are good or bad.
We'll be straight about our own product here. Fezerli writes drafts you post yourself, and it can also comment for you on autopilot. That part is automation. It's off until you switch it on, it types at a person's speed and only runs in the LinkedIn tab in front of you, but it still carries risk, and we say so on our homepage.
What makes a comment look like AI slop?
LinkedIn hasn't published a definition, and Srinivasan admitted it doesn't have a fixed one. Readers do, though. These are the tells people mention when they explain why they reported something, plus the ones we strip out of every draft we write.
| The tell | Typical example | Why it reads as AI | What a human would write |
|---|---|---|---|
| Empty praise | “Great post, thanks for sharing!” | Fits under any post ever written, so it proves you didn't read this one | Name the one line that stuck with you |
| Restating the post | “You're absolutely right that consistency is the key to growth.” | Adds nothing the author didn't already say | Add your own data point or story |
| The “not X, it's Y” flip | “It's not about posting more, it's about posting smarter.” | A favourite rhythm of language models in 2025 and 2026 | Just say the second half |
| Dashes everywhere | Two em dashes in three sentences | People type commas on a phone; models love the long dash | Commas, full stops, or nothing |
| Neat lists of three | “Clarity, consistency and courage.” | Too tidy for a quick reply | One idea, said once |
| Generic closing question | “What are your thoughts on this?” | The author wrote the post; they've shared their thoughts | Ask about something specific in the post |
| Invented numbers | “Studies show 73% of leaders agree.” | No source, suspiciously round, easy to check | Your own number, or no number |
| Missing the image | Praising “the insights” on a post that's one chart | The tool read the caption and skipped the picture | React to what the chart shows |
| Same shape every time | Compliment, paraphrase, question, on every post you touch | Visible the moment someone opens your activity tab | Vary length, tone and whether you ask anything at all |
| Wrong temperature | Upbeat emoji under a layoff post | Tone didn't match the room | Read the room, or skip the post |
The last two rows matter most, and they're the ones people forget. A single generic comment is forgettable. Twenty of them in your activity tab, all built the same way, is a profile people report.
Three rewrites, before and after
Great insights! Pricing strategy is so important for SaaS growth. Thanks for sharing your journey. What are your thoughts on freemium models going forward?
Going from 6% to 2.5% just by removing the free plan is wild. Did the people who would have churned stop signing up, or did they convert and stick around?
You're absolutely right, communication is key in the hiring process. It's not about the offer, it's about the experience. Respect goes both ways!
Ghosting after a final round stings. In fairness, I've also waited three weeks for a “we'll be in touch” that never came. Feels like both sides stopped believing the other will reply.
Congratulations on your new role! Wishing you all the best in this exciting new chapter. 🚀🎉
Eight months is a long search, and you posted the whole way through it. Congrats, they got someone who doesn't quit.
None of the good versions are clever. They're specific. That's the whole trick, and it's also why they're hard to fake at scale.
How do you use AI for comments without getting flagged?
Treat the AI as a fast first draft, never as the author. Here's what that looks like in practice for people who comment every day.
It starts with reading the whole post. Half of all slop comes from tools that see the first two lines and nothing else. If the post is a carousel, a PDF or a screenshot of a spreadsheet, the comment has to be about that. A tool that can't read images will praise a chart it never saw, and readers notice.
Then it needs your voice, not a house style. A model that has seen ten of your real comments, knows you never use exclamation marks and that you work in procurement will write something you'd actually say. One that hasn't will write like everyone else's tool, which is the exact sameness people report.
Edit one thing before you post. Usually it's a phrase that's a little too polished, or a question that's too broad. Swapping it for a detail only you know (a client story, a number from your own work) turns a draft into your comment. It takes ten seconds.
Keep it short. Most good comments are two or three sentences, well under LinkedIn's 1,250 character limit. Length is one of the easiest slop signals to spot, because a real person replying on a phone rarely writes four paragraphs.
Comment at a human pace. Fifty comments in an hour, all at 9:00? That looks like a bot because it is one. Spread them out, skip posts you've got nothing to say about, and leave the sensitive ones (layoffs, grief, politics) alone unless you genuinely know the person.
Reads the whole post, images included. Writes from your own examples. Strips the tells above. Waits for you to press Post. 50 drafts free, no signup.
What does this mean for comment bots and automation tools?
Bad news if your plan was volume. LinkedIn says it's already blocking hundreds of thousands of automated slop comment attempts a day, and billions of other automation attempts over the couple of months before the announcement. The slop button adds human reports on top of that detection. A comment bot now has two ways to get caught. The classifier, and the reader.
Engagement pods have the same problem. Twenty people from the same pod leaving “Great post!” within ten minutes is about the most reportable pattern there is, and every one of those reports trains the model that ranks the next post.
So is automation dead? Not entirely, and we'd be hypocrites to say so since we ship an Autopilot. But the bar moved. Automated comments that read as generic are now worse than no comments, because they invite reports. If you automate anything, automate only what you'd be happy to have written by hand, at a pace a person could keep, and expect that LinkedIn's rules still treat it as automation.
FAQ
Can I see who flagged my post or comment as AI slop?
No. LinkedIn treats it as a private report, like other reports. It hasn't announced any way to see who flagged you or how many times.
Does an AI slop flag reduce my reach?
LinkedIn hasn't published any penalty or threshold. A single flag hides the item for the person who reported it. The reports also train the classifiers that filter suggested content, so a steady pattern of flags is unlikely to help your reach, even though nobody outside LinkedIn knows the exact effect.
Can comments be flagged, or only posts?
Both. Hari Srinivasan's announcement mentions posts and comments, and a LinkedIn spokesperson said the option is available on all posts and comments.
When did LinkedIn add the “Seems like AI slop” button?
LinkedIn announced it on 30 July 2026 and described it as a test it was ramping up, so it reached accounts gradually over the following weeks.
Is it against LinkedIn's rules to use AI to write comments?
Writing with AI help isn't banned, and LinkedIn's product chief said people who use AI to refine their thoughts should keep doing so. Posting through bots or unauthorized automation is a different matter: section 8.2 of LinkedIn's User Agreement forbids automated commenting, liking and sharing.
How does LinkedIn detect AI generated comments?
It hasn't explained the technical side. LinkedIn says it uses new classifiers for low quality and AI slop content, blocks automated comment attempts, and uses the slop reports to tune those models.
What did LinkedIn replace “Enhance your post” with?
A proofreading tool. Srinivasan said it fixes mistakes without changing your voice, which is a fair summary of where LinkedIn now stands on AI writing.
Checked on 12 October 2026 against LinkedIn's announcement and coverage in TechCrunch, Fortune and Social Media Today (30 and 31 July 2026), plus LinkedIn's help page on prohibited software. LinkedIn changes features without notice, so tell us at admin@fezerli.com if something here goes out of date.