LinkedIn intensifies its efforts to combat low-quality AI content, achieving a 94% accuracy rate in detection.

      LinkedIn is taking action against AI-generated "slop" by reducing the visibility of generic posts in recommendations instead of removing them altogether. The platform claims a 94 percent accuracy rate in detecting such content in early tests, but has not provided data on false positives.

      If your LinkedIn feed seems like it is written by one individual with various accounts, you are not imagining things. The site has become a breeding ground for AI-generated posts that lack substance while sounding somewhat motivational. LinkedIn has announced measures to address this issue.

      The company is implementing changes aimed at what it terms "AI slop," referring to low-effort, AI-generated content that may appear refined but lacks genuine insight or expertise. Laura Lorenzetti, VP of Product, mentioned that the platform is developing detection systems to differentiate between posts that provide authentic viewpoints and those that seem repetitive, generic, and devoid of content.

      In initial tests, LinkedIn states its system successfully identified generic content 94 percent of the time. However, flagged posts will not be deleted; they will simply be hidden from recommendations. This means they remain visible to a user's immediate connections but will no longer circulate widely across the feed.

      The focus is precise. LinkedIn aims to combat clear engagement bait, recycled "thought leadership" that lacks originality, and posts demonstrating evident AI construction patterns, singling out the "it's not X, it's Y" formula as an example of the type of AI content to be downgraded.

      This crackdown will also apply to comments. LinkedIn plans to target bot-generated and generic AI comments that contribute nothing to discussions, often resembling a ChatGPT summary of the original post they respond to. The platform is also addressing automation tools that produce AI-generated content at scale.

      Nevertheless, a clear distinction is being made. LinkedIn asserts that AI-assisted content is still permitted if it includes original ideas or fosters meaningful discussion. The message is not to cease using AI, but rather to avoid allowing AI to take over the thought process entirely.

      Enforcing this distinction consistently will be challenging. While LinkedIn's claim of 94 percent accuracy is notable, the company has not disclosed data on false positives, leaving the frequency of legitimate posts being incorrectly flagged as slop unclear. Furthermore, LinkedIn has not specified how quickly these changes will be implemented, only noting that it may take months before users notice less low-quality AI content in their feeds.

      This action coincides with a growing emphasis on AI-generated content detection within the tech industry. OpenAI has recently adopted C2PA metadata and SynthID watermarks for its images, while ByteDance has introduced watermarking and IP protections in Seedance 2.0. However, text is considerably more difficult to label than images, and LinkedIn's strategy—relying on behavioral signals and stylistic cues rather than watermarks—is inherently less precise.

      The irony is evident, as LinkedIn is owned by Microsoft, a major investor in OpenAI, whose tools generate much of the content LinkedIn now seeks to limit. The platform also offers its own AI writing assistant, which automatically creates post drafts and comment suggestions, effectively building both the source and the filter simultaneously.

      Nonetheless, the need for AI-driven content moderation must begin somewhere, as LinkedIn's feed issue is real and worsening. If suppression proves effective, other platforms may follow suit. Conversely, if it fails, the company will publicly acknowledge that AI has compromised its feed without successfully correcting the issue.