LinkedIn’s attempt to remedy its AI slop problem started with a small button. Under the three dot menu on every post, you can now click “Seems like AI slop” and the platform hides it from your feed and quietly logs it as training data. It sounds minor. It is not. For anyone who has spent the last few years watching LinkedIn turn into a content landfill, this is the first real sign that the platform’s own numbers finally scared someone into action.
I want to walk you through what happened, why it happened, and whether it actually fixes anything. Not as a marketer trying to sell you a growth hack. As someone who runs an actual business, watches this platform closely, and has been annoyed by it for a long time.
Why LinkedIn Needed to Remedy Its Own Mess
Here is the plain truth. Over the past five years LinkedIn stopped being a place where professionals shared real experience and slowly turned into a stage for performance. You know the posts. “I lost everything. Here’s what it taught me about B2B sales.” A tragedy, a lesson, three bullet points, and a call to book a consult. Every week, a new version of the same formula.
The data backs up what anyone active on the platform already felt in their gut. AI detection firm Pangram found that more than 40 percent of long form LinkedIn posts are now fully AI generated, and LinkedIn accounts for roughly 62 percent of all AI generated content the firm scanned across major social networks. That is not a rounding error. That is a platform where nearly half of what you scroll past was never written by the person whose name is on it.
And it did not happen by accident. When platforms optimize for the fastest engagement signals, clicks, quick likes, easy comments, the system gets gamed by whoever can produce the most content the fastest. That is not new. Newspapers figured this out over a century ago with the “if it bleeds, it leads” model of yellow journalism. Whatever triggers the strongest immediate reaction wins the front page, regardless of whether it deserves to be there. LinkedIn just built a faster, more automated version of the same trap, and generative AI gave everyone a printing press.
The result was predictable. People with no real business experience became the loudest voices on the platform, some with hundreds of thousands of followers, while people actually doing the work, running companies, fixing operations, managing real clients, got buried under the noise. Real operators looked at the fake vulnerability hooks and the engagement pods, groups of fifty creators liking each other’s posts within ten minutes of publishing, and many of them just walked away. You cannot really blame them. If you refused to play the game, your honest post sat at twelve views while the AI slop next to it hit fifty thousand impressions.
What LinkedIn’s Attempt to Remedy the Problem Actually Looks Like
On July 30, 2026, LinkedIn rolled out the “Seems like AI slop” flag. Chief Product Officer Hari Srinivasan called it a top priority for the company and said the flags feed directly into the platform’s detection models. His own words were blunt about the difficulty of the task. Slop is hard to define and the definition keeps changing, so the flagging data helps the models keep tuning what they catch.
A few other pieces moved at the same time, and together they tell you LinkedIn is serious rather than just doing a PR gesture.
- LinkedIn removed its own “Enhance your post” AI writing tool, the one that used to rewrite your wording for you, and replaced it with a proofreading feature that fixes grammar without changing your voice. LinkedIn building AI slop with one hand while flagging it with the other was never going to hold up, so they killed the tool that was contributing to the problem.
- There is now a private nudge inside your own analytics dashboard. If people flag your content, LinkedIn quietly tells you your post came across as inauthentic or AI heavy. It is not a public penalty. It is feedback meant for you alone.
- The platform says it is blocking billions of automated comment and engagement attempts, on top of the new flagging system.
None of this exists in a vacuum either. Since 2024, LinkedIn has been quietly replacing its old patchwork of ranking systems with a single model called 360Brew, a 150 billion parameter foundation model that reads posts, profiles, and job history the way a language model reads any other text, instead of relying on old fashioned click and like counters. The full rollout across the feed landed in March 2026, and the effect on reach has been dramatic. Median post reach is down somewhere around 47 to 50 percent, video reach has dropped as much as 72 percent in some cases, and quality signals like saves and slow, thoughtful engagement now carry four to six times more weight than a simple like.
Put plainly, LinkedIn shifted from a Social Graph, who you know and who likes your post in the first ten minutes, to something closer to an Interest and Depth Graph, what you actually know and whether real people stop scrolling long enough to read it. The flagging button is the public facing part of that shift. 360Brew is the engine underneath it.
The Psychology Behind Why This Took Five Years
If you are wondering why it took LinkedIn this long to act, there is a simple explanation, and it is the same one that explains a lot of bad platform design across the internet. Variable, fast engagement is addictive to measure and easy to reward. A like takes half a second. A save, a returning reader, a genuine multi sentence comment from someone who actually works in your field, that takes longer to earn and longer to notice in the data.
This is the same dynamic B.F. Skinner described with intermittent reinforcement almost a century ago, minus the AI. Unpredictable, fast rewards keep people pulling the lever. Platforms built their early recommendation systems around exactly that kind of reward because it was measurable and it drove time on site. The problem is that when you reward the lever pull instead of the value behind it, you eventually get a feed full of people who are very good at pulling levers and not very good at anything else.
There is also a simpler economic idea at play here, the one behind Gresham’s Law, that bad currency drives good currency out of circulation once people realize the bad version is easier to produce and just as accepted. Apply that to content. Once AI generated posts could be produced in seconds and rewarded the same as a genuine breakdown that took a professional two hours to write, the honest content got pushed out of circulation. Not because it was worse. Because it could not compete on volume.
LinkedIn’s numbers eventually had to reflect that imbalance. Users complaining publicly, a media outlet like 404 Media reporting directly on the flood of AI slop, and even outside voices like Substack’s CEO citing LinkedIn as a warning sign of what happens when a platform lets synthetic content run unchecked, all of that pressure adds up. A company does not remove its own AI writing tool and ship a flagging button as a “top priority” unless the trust numbers were flashing red internally.
What LinkedIn’s Attempt to Remedy AI Slop Means If You Actually Run a Business
Here is where this stops being industry gossip and starts mattering for anyone trying to build real authority and real clients on the platform.
Proof of work now beats generic advice. Posting “3 ways to improve operational efficiency” gets buried. Posting the actual story of how you restructured a client’s intake workflow from four days down to three hours, including where it nearly broke, gets read. Generic AI content can summarize a principle in seconds. It cannot fabricate your specific mistakes, your specific fix, or your specific numbers.
Saves and dwell time matter more than likes. A like is nearly worthless now, heavily discounted by the ranking model because it is trivially easy to fake. A save signals someone thinks your post is worth coming back to. A person reading for thirty seconds or longer signals they are actually digesting what you wrote, not scrolling past it.
Your profile has to match your content. The model reads your headline, About section, and Experience history as context before deciding how far to push your posts. If your headline is vague and your posts are about something completely different, the system has a harder time deciding who should see your work.
Direct, personal relationship building now outperforms broadcasting. Posting five times a week with shallow content loses to posting two or three times with real depth, paired with spending real time writing thoughtful comments on posts from the people you actually want as clients or peers. A three sentence comment that debates a specific detail of someone’s post carries more weight in this system than twenty “Great post!” replies ever will, and it usually starts an actual conversation instead of a fake one.
None of this is a trick. It is closer to how professional reputation worked before social platforms existed at all. You did good work, people who saw the work vouched for you, and your reputation spread slowly through people who actually knew what you were talking about. LinkedIn did not invent this. It just spent five years rewarding the opposite of it, and it is now trying, imperfectly, to remedy that.
Is LinkedIn’s Attempt to Remedy This Actually Going to Work
I want to be honest instead of optimistic for the sake of it. Detecting AI generated text reliably is genuinely hard, and LinkedIn has said as much itself. The definition of slop keeps shifting, flagging systems can be abused by people who just do not like a post’s opinion, and a private nudge in someone’s analytics is not going to change five years of habits overnight.
But the direction is right, and the incentives finally point the correct way for the first time in years. A platform that rewards saved posts, real dwell time, and profile consistent expertise over quick likes is a platform where showing your actual work outperforms performing for an audience. That is a genuinely positive shift for people doing real work in food operations, consulting, or any other field where the value is in the details, not the hook.
The slop is not gone. But for the first time in a long while, the platform’s own math is finally on the side of the people who never stopped doing the work.
