Scroll through LinkedIn long enough and you will eventually hit one of these posts. It opens with a bold personal claim, pivots into a tidy life lesson, and lands on a line that reads something like “it’s not about the title, it’s about the impact.” The post gets a few hundred reactions from accounts you have never heard of, and you scroll past it without remembering a single word five minutes later. That specific rhythm has a name now, and LinkedIn itself is the one giving it that name: AI slop.
The company detailed a new detection and suppression system this year aimed squarely at that kind of content, and the rollout says a lot about how far the problem has actually spread across the platform that built its identity on professional credibility.
What LinkedIn Is Actually Rolling Out
According to Laura Lorenzetti, LinkedIn’s VP of Product, engineers spent months working alongside the platform’s editorial team to identify patterns that separate a post with genuine perspective from one that just repeats an idea everyone has already read a hundred times. In early testing, the resulting system correctly flagged generic, low value content 94 percent of the time.
Flagged posts are not deleted. They are suppressed from recommendations, meaning a poster’s direct connections and followers can still see the content, but it stops spreading further into the wider feed. That distinction matters because it lets LinkedIn sidestep a much harder fight over free expression while still starving the content of the oxygen it needs to go viral.
The system is trained to catch more than one flavor of slop. It targets outright engagement bait, recycled “thought leadership” that adds nothing new, and comments that show obvious tells of machine authorship, including the now familiar “it’s not X, it’s Y” contrastive phrasing that shows up in an enormous share of AI generated text. LinkedIn is also going after what it calls attention bait videos: long clips of construction accidents paired with vague workplace safety commentary, or extended manufacturing footage set to generic business platitudes, all designed purely to keep a thumb from scrolling rather than to teach anyone anything.
The Scale of the Problem
The scope of what LinkedIn is dealing with becomes clearer once you look at independent research on the topic. A study from AI detection firm Pangram found that roughly 41 percent of longform LinkedIn posts, meaning anything over 250 words, now show signs of AI generation. That is not a fringe behavior from a handful of spammy accounts. It is close to half of the platform’s substantial written content.
The trend is not unique to LinkedIn either. Coverage from outlets tracking the broader shift has pointed to a similar pattern spreading across X and other text heavy platforms, where a small set of AI writing habits, from the contrastive sentence structure to a fondness for tidy three item lists, have become so common that readers now spot them almost instinctively, the same way people learned to recognize a chain email years ago.
Where the Line Gets Blurry
LinkedIn has been careful to frame this as a crackdown on low value content rather than a ban on AI assisted writing itself. That distinction is easy to state and much harder to enforce. A founder who uses an AI tool to tighten up a post built around a real story from their own week is doing something meaningfully different from someone who prompts a chatbot to generate a motivational paragraph out of thin air, but both posts can end up looking similar on the surface, especially to an automated classifier scanning for sentence patterns rather than substance.
That tension is really the whole story here. A platform cannot easily build a filter for “does this person actually have something to say” without occasionally punishing well crafted, AI assisted writing alongside the genuinely empty kind. LinkedIn’s bet is that suppression rather than removal gives it enough room to be wrong sometimes without doing lasting damage to legitimate voices caught in the net.
Part of a Bigger Pattern in How AI Is Reshaping Content
This crackdown lands at an interesting moment for the wider conversation about AI and online content. On one side of the industry, publishers are watching AI chatbots quietly reroute how people find information online, pulling clicks away from the sites that did the original reporting. On the other side, a platform like LinkedIn is fighting the opposite problem: too much AI generated content flooding a feed that was never designed to filter out machine written filler at this kind of volume. Both stories are really about the same underlying shift, just approached from opposite ends of the pipe that carries information from a source to a reader.
There is also an economic angle that helps explain why the slop problem grew so fast in the first place. Even as companies have started questioning whether AI compute is actually cheaper than paying a human to do the same job, generating a single social media post costs next to nothing by comparison. That lopsided economics, expensive enterprise AI on one hand and essentially free content generation on the other, is exactly the kind of gap that floods a platform with low effort posts long before anyone builds a system to catch them.
What This Means If You Actually Post on LinkedIn
For anyone who uses the platform regularly, the practical takeaway is fairly simple. Posts built around a specific, verifiable detail from your own work tend to read as authentic almost by default, because generic AI text has a hard time faking specificity convincingly. Vague inspirational framing, the “it’s not X, it’s Y” construction, and paragraphs that could have been written about literally any industry are the patterns most likely to get quietly buried.
None of this means AI writing tools are the enemy here. Writing assistants themselves keep getting better at sharpening a person’s actual voice rather than replacing it outright, and LinkedIn has said repeatedly that AI assisted posts are welcome as long as they carry a real point of view. The tools are not the problem. Using them to manufacture the appearance of insight where none exists is.
The Bigger Picture
Whether this actually works at scale is still an open question. A 94 percent detection rate in a controlled test is one thing. Holding that accuracy against millions of posts a day, written by people actively trying to sound less like a machine once they realize what is being filtered, is a much harder and more permanent fight. Content moderation systems have a long history of arms races between platforms and the people trying to game them, and there is no obvious reason AI slop detection will be the exception.
What the announcement really confirms is something anyone who spends time on the platform already suspected. The feed had a quality problem large enough that LinkedIn felt compelled to build a dedicated system to fight it, rather than trusting its existing ranking signals to sort things out on their own. That alone tells you how much of what shows up in a professional network’s feed today was never written by a professional at all.
