LinkedIn has shipped a button that lets you report a post for seeming like AI slop. It sits in the three-dots menu, and The Verge reports that clicking it hides the post and thanks you for the feedback. Chief product officer Hari Srinivasan says "AI slop is a top priority for all of us" and that the reports will help the company "tune our models and make better feeds."
Since Facebook, X, and Reddit all feel like they're drowning in AI slop these days, should these other platforms also do the same thing?
Why LinkedIn shipped this
A study from AI detector Pangram, which scanned 1,002,627 posts through a browser extension starting in April, found more than 40 percent of LinkedIn longform posts flagged as fully AI-generated. LinkedIn alone accounted for 62 percent of all the AI content the study flagged anywhere.
Nearly two-thirds of the writing is done by AI, in one place, on the platform people join for work and networking.
The platforms are not equally bad, and Reddit is the surprise
Here's where the answer to "should everyone add this" gets complicated. The same study puts X at 23.9 percent fully AI-generated posts with another 22.9 percent AI-assisted, leaving barely half of the posts fully human. Substack came in at 21.9 percent.
Reddit was 4.4 percent. Replies on Reddit were 98.1 percent human-authored.
That contradicts what I would have guessed based on anecdotal evidence, and I've written more than once about Reddit feeling full of bots. Some of that gap can be explained by measurement, since the study only looks at writing of 50 words or more and a lot of Reddit is fewer words and a joke. But a 4.4 percent AI share against LinkedIn's 40-plus is a huge margin and it points at something structural.
Reddit already has the button. It's called a downvote.
Look into what Reddit has that LinkedIn doesn't, and the answer is pretty clear. Reddit has downvotes, so a low-effort post gets buried by the people reading it, in seconds, at no cost to anyone. Reddit has human moderators who remove posts and ban accounts. Reddit has communities whose regulars notice a stranger talking like ChatGPT.
LinkedIn has none of that. There's no downvote. There's no meaningful moderation of tone. And the incentive runs exactly the wrong way, because posting on LinkedIn is career-adjacent, so people post whether or not they have anything to say, and the safest way to produce something on a day you have nothing is to ask an LLM model for it.
Which makes the slop button a patch on a design problem. It's a way of bolting a downvote onto a platform that deliberately never had one, except it's routed through a reporting flow and a classifier instead of just being a downvote.
The hidden part of the announcement
LinkedIn is removing its own AI feature that "enhanced" your post, and replacing it with one that proofreads without changing your voice.
That's LinkedIn admitting it built part of the machine it's now fighting. Handing every user a one-click "make this sound more professional" tool, on a network where everyone is performing professionalism, was always going to produce that AI-slop voice.
So should the other platforms follow suit?
X, yes, and for the same reason as LinkedIn: it has an engagement payout, an incentive to bait and get replies economy, and no downvote. When barely half of long posts are fully human-written, a reader-side signal is worth doing.
Reddit, no for now. It already has what the button is imitating, and running a classifier over a site that's 98 percent human replies would produce mostly false accusations. Facebook is its own problem, because the slop there is AI slop generated images and engagement farms rather than longform text, and a "seems like AI slop" button maybe could help. If enough people report “seems like AI slop” on Facebook posts maybe my feed will actually be cleaned up.
What stays unsolved
Pangram's own CEO Max Spero called the figures a lower bound. Detection is a moving target, the people running slop at scale will tune around any classifier that gets deployed, and the genuine users are the ones who'll get wrongly flagged along the way.
LinkedIn is also testing a way to tell people when others think their posts read as inauthentic. I have no idea whether that shames anyone into writing their own posts or just teaches them to launder the output better. Ask me in six months.
Sources
The Verge: LinkedIn actually adds a 'seems like AI slop' button - Jay Peters on the new report button in the three-dots menu, the hide-and-thank behaviour, Hari Srinivasan's comments, the new classifiers for AI-slop and low-quality content, the test that tells posters their content reads as inauthentic, and the removal of the AI post-enhancement feature in favour of a proofreader that does not change your voice.
Pangram: AI content is everywhere on social media - The study of 1,002,627 posts scanned from April 2026, with LinkedIn longform above 40 percent fully AI-generated and accounting for 62 percent of all flagged AI content, X at 23.9 percent fully AI and 22.9 percent AI-assisted, Substack at 21.9 percent, Reddit at a 4.4 percent combined share with 98.1 percent human replies, the 50-word minimum, and the lower-bound framing.
404 Media: LinkedIn and X are flooded with AI spam, browsing data suggests - The original reporting on the browsing-extension methodology behind the platform figures.
