LinkedIn Cracks Down on AI Slop with 94% Detection Accuracy
LinkedIn is finally cracking down on AI-generated posts, often referred to as "slop," and the feed might actually become readable again. The Microsoft-owned platform will suppress generic AI-created content rather than removing it, claiming 94% accuracy in early detection tests.
TL;DR
LinkedIn is tackling AI-generated “slop” by suppressing generic posts from recommendations, instead of removing them. Their system, trained to recognize repetitive and empty content, flags such posts with 94% accuracy. Suppressed posts will still be visible to direct connections but won’t spread across the wider feed.
If your LinkedIn feed felt like it was written by one person managing ten million accounts, you weren’t imagining things. The platform has become a breeding ground for AI-generated posts that sound inspirational but offer little original content. Now, LinkedIn says they are taking measures to fix this.
Changes and Targets
The company announced changes aimed at "AI slop"—low-effort, AI-generated content that lacks originality and meaningful thought. They specifically target outright engagement bait, recycled "thought leadership," and formulaic posts with obvious AI construction patterns, such as "it’s not X, it’s Y."
This crackdown extends to comments too, focusing on bot-generated or generic AI comments that add nothing to conversations. LinkedIn is also targeting tools that generate AI content at scale through automation.
However, LinkedIn draws a line between AI-assisted and AI-generated content. They welcome the latter as long as it contains original ideas or encourages meaningful dialogue. The focus is not on stopping AI usage but on ensuring humans remain involved in the thought process.
Challenges and Unanswered Questions
While LinkedIn’s 94% accuracy claim sounds impressive, they have not shared data on false positives. It remains uncertain how often legitimate posts might be wrongly flagged as slop. Additionally, the rollout timeline is unclear; it could take several months before users see reduced low-quality AI content in their feeds.
This move aligns with growing priorities across the tech industry for AI-generated content detection. OpenAI and ByteDance have implemented various measures to fingerprint and watermark their AI outputs, but text detection poses unique challenges compared to images.