LinkedIn drowning in AI slop as 41% of longform posts flagged as fully AI-generated

AI slop writing has taken over the internet, particularly LinkedIn and X

No surprise here. A study from AI detection platform Pangram suggests that social media posts are teeming with AI-generated slop, particularly if the posts are long and especially if they live on LinkedIn or X. If you're sick of reading non-human prose, we'd recommend getting off the platforms altogether. Along with offering your typical AI-content detection services, Pangram released a Chrome extension at the end of April that, with a $20/month subscription, will automatically scan a user's LinkedIn, Medium, Substack, X, and Reddit feeds to check for AI-generated or assisted content. With more than one million posts analyzed from users who opted in to share data through the extension since its launch, Pangram has concluded that, while AI slop is flooding social media, it's hitting longform content particularly hard. With longform content defined in its study as any post over 250 words, Pangram found that a full 25 percent of such posts across all the platforms it studies were fully AI-generated. Fully, mind you, meaning that doesn't include posts in which users got the assistance of an LLM to gussy up their bland prose. That average across platforms was hardly evenly distributed, though. Leading the way was LinkedIn, where 41 percent of longform content was fingered by Pangram as being AI-generated. That's likely unsurprising to anyone who's ever bothered to read a lengthy professional diatribe from the Microsoft-owned slop shop, or for El Reg readers - a prior story we reported on in late 2024 from AI detection outfit Originality.ai found that 54 percent of LinkedIn longforms were AI-generated. Originality's definition of Longform was a bit looser, however, with anything over 100 words counting in its analysis. Per Pangram, shortform content on LinkedIn isn't much more likely to be human authored - they found 30 percent of posts between 50 and 250 words were fully written by AI. For LinkedIn thought slop leaders, it's generally all or nothing when it comes to using AI to write posts, with a mere 4.3 percent of longform content written with AI assistance. On the other hand, only 55.2 percent of longform posts on the platform, Pangram concluded, are actually written by humans. While LinkedIn may take the cake in terms of the volume of full-slop longform posts, Elon's X has it beat when adding partially-written AI garbage into the mix, but not by much, honestly. A quarter of posts on X are fully AI authored, and an additional 23.2 percent are believed to be written with AI help. That leaves 52.7 percent of Twitter posts attributed to humans. In effect, you're roughly batting .500 on either site. Pangram found that Medium isn't that much better, with roughly one in three posts likely to have been written by, or with the aid of, an AI. Substack was far and away the least likely place to find AI slop in disguise, but even then, nearly a quarter (21.9 percent) of posts analyzed by the Chrome extension were written by or with AI. Reddit is a slightly more complicated situation, with comments on posts making up a large portion of Reddit content. According to Pangram, 11.6 percent of Reddit posts are AI authored or assisted; 98.1 percent of comments were found to be human authored, and the sheer quantity of comments vs. top-level posts meant that Reddit appears to be the place to go if you want to avoid an intrusion of AI thinking. All said, Pangram concluded from its data that AI writing is flooding social media, just like it's flooding websites and basically everywhere else online. "An internet that is completely flooded with undisclosed AI content is bleak, but we don't believe it's inevitable," Pangram CEO Max Spero said of his company's findings in the report. Pangram believes letting internet users know what's been AI-generated so they can ignore it is a solution to the problem, but you'll have to pay $20/month if you want the Chrome extension to provide that service. It's still usable without paying, but content has to be manually input, and the daily limit is just 4,000 words. In other words, unless you want to pony up and see who's bullshitting you on social media, you'll have to just assume everyone is. Like we suggested up top, maybe it's time to disconnect from those feeds entirely. ®

I knew there was plenty of AI slop on LinkedIn. Shocking report says the problem is far worse than suspected

I already knew LinkedIn was overflowing with posts written by AI, recycled leadership advice, and those god-awful lessons about entrepreneurship. A new report suggests the situation is considerably worse than even the platform's feed makes it appear. AI-detection company Pangram analyzed more than one million posts scanned through its Chrome extension across LinkedIn, X, Reddit, Medium, and Substack. LinkedIn represented approximately one-third of everything scanned, yet produced 62% of all content Pangram flagged as AI-generated. LinkedIn is leading the AI slop race Across every platform included in the dataset, 13.8% of scanned material was classified as fully AI-generated. The proportion climbed sharply as posts became longer, reaching 25.72% among items containing more than 250 words. LinkedIn performed especially poorly. More than 40% of its long-form posts were flagged as fully AI-generated, the highest rate recorded across the five platforms. A top-level LinkedIn post was also 1.35 times more likely to be classified as AI-generated than a comment. Recommended Videos Even the comment section did not escape this. After accounting for length, LinkedIn comments were slightly more likely to contain AI writing than posts. X also produced an unpleasant result. Pangram classified 23.9% of long-form X articles as fully AI-generated and another 22.9% as containing a mixture of human and AI writing. Only 53.2% were judged fully human-authored. The numbers come with some important caveats Pangram gathered the data from people who installed its extension and voluntarily shared anonymous scanning statistics. This makes it a large convenience sample rather than a random, representative snapshot of everything published on LinkedIn. It also relies on Pangram's own model, which the company claims has a false-positive rate of 0.01%. Although no detector can establish authorship with absolute certainty. LinkedIn acknowledged its slop problem in May and said it would reduce the recommendation reach of repetitive AI content. The company claimed its early system correctly identified generic material 94% of the time.