LinkedIn’s new “seems like AI slop” button lets users flag low-quality AI-generated content, and it’s already been used over a million times. But LinkedIn has never actually defined what counts as slop, and neither has anyone else. The button measures a reaction, not a standard. Until organizations and platforms agree on what separates “AI-touched” from “AI slop,” a lot of genuinely useful, expert-reviewed content is going to get flagged right alongside the spam.
In late July 2026, LinkedIn rolled out a new reporting option inside the post’s three-dot menu, alongside the standard “report post” choice: a “seems like AI slop” flag, built to let users call out low-quality, obviously AI-generated content in their feed. LinkedIn says the signal helps tune its classifiers, and posts that get flagged enough start seeing meaningfully reduced reach: the company has reported roughly 40% fewer views on content its systems tag as low-quality AI output. Users have clicked the button over a million times since launch.
What LinkedIn hasn’t done is say, in plain terms, what “AI slop” actually means. That’s the gap Merav Yuravlivker, Chief Learning Officer and Co-Founder of Data Society, zeroed in on:
“The AI slop button on LinkedIn is an interesting use case of how we want to move forward as a society around AI on the internet, because AI touches many pieces of content. Does that count as AI slop? What is the definition that people have in their heads around this term, slop versus not?”
Merav Yuravlivker, Chief Learning Officer and Co-Founder, Data Society
That’s not a small omission. Third-party estimates suggest AI now touches a substantial share of longform LinkedIn posts, with one 2026 analysis of public posts finding over 80%. If “AI touched it” were the bar, most of the platform’s content would qualify as slop. Clearly that’s not what anyone means. But nobody, including LinkedIn, has said what the actual bar is.
In the absence of an official definition, Yuravlivker offers her own, and it isn’t about the tool at all:
“To me, if a piece of content is helpful and it’s accurate and it’s overseen by an expert, whether it’s touched by AI personally, it’s just not that important to me, because I know that it’s been vetted well by a human. If somebody is releasing a bunch of spam, there’s also a report post button on LinkedIn as well, and I’m curious about how LinkedIn has thought through what the difference is between reporting the post versus reporting it as AI slop, and how they would categorize that.”
That reframes the question usefully. The relevant variable isn’t whether AI was involved in drafting a piece of content. It’s whether a qualified human reviewed it before it went out, and whether it’s actually useful once it did. A post written entirely by a human but full of bad information isn’t better than an AI-assisted post that’s accurate and expert-reviewed. Under Yuravlivker’s framework, the first is arguably worse, it just doesn’t have a button for it.
That also surfaces a real design question LinkedIn hasn’t publicly answered: what’s the actual difference between the existing “report post” option (built for spam, harassment, and policy violations) and the new “AI slop” flag? If both funnel into similar enforcement, the new button may mostly be adding a label to behavior that was already reportable. If they trigger different consequences, users deserve to know what those are.
Yuravlivker raises a point that’s easy to miss: the system evaluating whether something is “AI slop” is, itself, almost certainly AI.
“Let’s not kid ourselves, I’m sure that LinkedIn, regardless of which button you use, is using AI to review those posts. I don’t know exactly what the consequences are around reporting it as AI slop. Does that user get a verbal warning? I think people are going to use it pretty liberally at first.”
LinkedIn has confirmed as much: the company has said it’s deploying automated classifiers to identify AI-generated and low-quality posts and reduce their visibility, and that it will privately flag inauthentic-seeming content in users’ own dashboards, framed as a nudge to improve their writing rather than a public penalty. What LinkedIn hasn’t detailed is exactly how a report translates into an outcome, whether one flag does anything, how many are needed, or what happens to a user’s account or reach as a result. That ambiguity is precisely what Yuravlivker predicts will drive early overuse: when people don’t know what a button does, they tend to press it on anything that feels off, rather than anything that meets a clear bar.
Yuravlivker’s conclusion is that this problem doesn’t have a clean technical fix:
“How can somebody say that something is AI slop versus something is generated by AI and supported with human oversight? It’s very difficult to find those lines and that balance, so we’ll see how we progress. I think that’s a human choice, and if human behavior is involved, we’re going to see some ups and downs before we come to a good equilibrium.”
For organizations publishing AI-assisted content, that’s the practical takeaway. The safest ground is demonstrable human oversight: expert review, fact-checking, editorial judgment behind whatever gets published, so that if a flag ever lands on your content, there’s a real answer for why it doesn’t belong there. Content that’s helpful, accurate, and vetted by someone who knows the subject has a defense. Content that’s neither AI-assisted nor reviewed carefully has none, regardless of which button gets clicked.
LinkedIn’s button measures something real: a rising tide of low-effort AI content nobody wants in their feed. But it’s a human problem before it’s a technical one. Data Society’s AI Advisory work helps organizations build the editorial oversight that stands up to a flag, and AI upskilling programs build the judgment behind responsible AI-assisted publishing.
