The amount of AI-generated content you scroll through might be far greater than you think, according to a new study by the AI detection platform Pangram.
Pangram asked users of its Chrome extension to opt in to sharing their browsing results with the platform. From those results, Pangram used its AI detection model—which has a 0.01% false positive rate, according to the company—to scan nearly 1 million posts that users scrolled through across the internet over the course of two months.
Pangram only scanned posts that users came across—meaning that AI slop is not just contained to spam sites but actually prevalent on the social media platforms that humans frequent most, including LinkedIn, X, Medium, Reddit, and Substack.
On X, a quarter of long-form content was fully written by AI, but another 23% was AI-assisted. Substack was the long-form platform with the lowest rate of AI-written content, yet more than one-fifth of its posts were still flagged as AI-generated or -assisted.
But LinkedIn had the highest proportion of AI-generated pieces, the Pangram data suggests.
More than 40% of LinkedIn posts longer than 250 words were flagged as being fully written by AI. And while LinkedIn made up for a third of Pangram’s scanned content, posts on the platform accounted for nearly two-thirds of all the AI content flagged.
Pangram further determined that top-level LinkedIn posts were 1.35 times more likely to be AI-generated than a comment. Still, LinkedIn comments were slightly more likely to be generated by AI compared with top-level posts on other platforms.
LinkedIn includes a built-in “Enhance post” button that offers users AI writing assistance on their posts, making the process of using AI simple and easy.
In response to a comment request from Fast Company, LinkedIn pointed to a May post from global editorial executive editor Laura Lorenzetti that announced the platform would detect and downrank content that appears to be AI-generated so users would see it less on their feeds.
“While AI can be a helpful tool for refining language, we’re seeing a rise in what many call ‘AI slop’—content that is low-effort, AI-generated content that may sound polished on the surface but lacks any real unique perspective or substance,” Lorenzetti wrote in the post. “When AI is overused, especially at scale and in an automated way, it dilutes the valuable insights that real human conversations can spark.”