LinkedIn adds 'AI slop' button as 41% of longform posts are flagged as bot-made

LinkedIn is finally giving you a way to narc on the AI-generated dreck clogging your feed. A new “Seems like AI slop” option is rolling out inside the three-dots menu on posts, letting users flag content that doesn’t read like a human wrote it. When you hit it, the post vanishes and you get a polite thank-you note. It’s a blunt instrument for a blunt problem. The move comes as AI detector Pangram found that 41 percent of longform LinkedIn posts were entirely machine-generated, according to 404Media. That’s not a typo — nearly half of the thought leadership you’re scrolling past might be from a prompt, not a person.

Chief product officer Hari Srinivasan didn’t mince words. “AI slop is a top priority for all of us,” he wrote. “We really care about this.” I believe him, mostly because the platform’s value proposition — authentic professional networking — dissolves when your feed becomes a landfill of ChatGPT summaries dressed up as career advice. The report button isn’t just for show, either. Srinivasan says the data will help “tune our models and make better feeds.” Translation: every time you flag a post, you’re training LinkedIn’s classifiers to spot the synthetic stuff before it reaches anyone else.

Under the hood, LinkedIn is deploying new automated classifiers that don’t wait for user reports. These systems will try to identify low-quality or AI-generated content proactively, particularly in suggested posts and content from outside your network. The company is also toying with a nudge feature that tells users when others think their posts feel inauthentic or AI-heavy. It’s a soft-shame approach — less scarlet letter, more passive-aggressive notification.

Perhaps the most telling change is what LinkedIn is killing. The platform is axing its own AI tool that would “enhance” your posts, the very feature that probably contributed to the slop flood in the first place. It’s being replaced by a proofreading tool that fixes your grammar but doesn’t rewrite your voice. That’s an admission that the cure was part of the disease. The question now is whether a report button and some classifiers can actually drain the swamp, or if we’re just giving users a tiny paddle while the flood keeps rising.