If your LinkedIn feed has felt like a sea of generic, inspiration-laced but ultimately hollow posts, you are not alone. The professional networking platform has become a breeding ground for AI-generated content that says nothing while pretending to offer valuable insights. Now, LinkedIn says it is taking decisive action to reclaim the feed's quality by targeting what it calls 'AI slop'.
In a significant policy shift, LinkedIn announced changes designed to detect and suppress low-effort, AI-generated posts that lack original thought or genuine expertise. According to Laura Lorenzetti, LinkedIn's Vice President of Product, the platform has developed detection systems trained to differentiate between content that adds real perspective and content that is repetitive, generic, and essentially empty. Early internal tests suggest the system is able to correctly identify such low-quality posts 94 per cent of the time, although the company has not released any data on false positives—a crucial metric for any content moderation tool.
The problem of AI slop
The rise of generative AI, especially large language models like OpenAI's GPT series, has made it incredibly easy to produce text that sounds plausible but lacks substance. On LinkedIn, this has manifested in a flood of posts that follow predictable patterns: 'It's not X, it's Y' formats, recycled 'thought leadership' clichés, and outright engagement bait. These posts often attract comments and likes, but they contribute little to meaningful professional discourse. Users have complained that their feeds have become monotonous, with many accounts essentially regurgitating the same shallow advice.
LinkedIn's crackdown is not a ban on AI-assisted writing. The company has made a deliberate distinction between content that is entirely AI-generated—where the AI does all the thinking—and content that uses AI as a tool to refine or assist human ideas. The latter is still welcome, provided it contains original insights or encourages genuine conversation. In practice, this line will be difficult to enforce consistently. The detection system relies on behavioural signals and stylistic patterns rather than technical watermarks, which introduces inherent fuzziness.
How the detection works
LinkedIn's approach to identifying AI slop focuses on multiple indicators. The system looks for repetitive structures, formulaic language, and engagement-bait tactics. It also analyses comments, targeting bot-generated or generic AI replies that add nothing to the conversation—for instance, comments that simply summarise the original post in a ChatGPT-like tone. Automation tools that generate AI content at scale are also in the crosshairs.
Flagged posts will not be removed outright. Instead, they will be suppressed from recommendation algorithms, meaning they will still appear to the poster's immediate connections but will not be amplified across the wider feed. This is a softer approach than deletion, but it effectively limits the spread of low-quality content. The system's accuracy is touted as 94 per cent, but without transparency on false positive rates, there is a risk that legitimate posts—especially those from non-native speakers or users who employ AI for assistance—could be inadvertently demoted.
The broader context of AI content moderation
LinkedIn's move comes at a time when the tech industry is grappling with the challenges of AI-generated content across all media. OpenAI has adopted C2PA metadata and SynthID watermarks for images generated by its DALL-E models. ByteDance added watermarking and IP guardrails to its Seedance 2.0 video generation platform. However, text is far more difficult to fingerprint than images or video. Watermarks are not practical for text, so platforms must rely on behavioural analysis and stylistic detection, which are inherently less precise.
The irony of LinkedIn's crackdown is not lost on observers. LinkedIn is owned by Microsoft, which is one of the largest investors in OpenAI—the company whose tools produce much of the content LinkedIn now wants to suppress. Moreover, LinkedIn itself offers an AI writing assistant that can auto-generate post drafts and comment suggestions. The platform is, in effect, simultaneously providing the firehose and building the filter. This dual role raises questions about the effectiveness and impartiality of the moderation system.
Potential impact and user reactions
For many users, the announcement is a welcome relief. The feed has become saturated with posts that feel like they were written by one person with ten million accounts. The crackdown could restore some credibility to the platform as a space for genuine professional exchange. However, there is uncertainty about the rollout timeline. LinkedIn has indicated that it could take several months before users see noticeably less low-quality AI material in their feeds. The company has not specified how the system will handle different languages or cultural contexts, where patterns of formulaic writing may differ.
There are also concerns about false positives. If the system is too aggressive, it could suppress posts from users who rely on AI to overcome language barriers or who use AI to structure their thoughts without losing originality. The lack of transparency on error rates means that users will have to trust LinkedIn's algorithm—an algorithm that has historically been opaque in its recommendations.
Comparison with other platforms
Other social networks are also struggling with AI-generated content. Facebook and Instagram have invested in similar detection tools, but with mixed results. Reddit has banned AI-generated content in certain subreddits, and Twitter (now X) has been criticised for allowing bots and spam to proliferate. LinkedIn's approach is notable because it explicitly targets the 'thought leadership' genre, which is central to its brand. By focusing on substance over style, LinkedIn hopes to differentiate itself as a platform where professional expertise is valued.
The success of this initiative will depend on continuous refinement. AI-generated text is evolving rapidly, and detection systems must adapt to new patterns. LinkedIn has the advantage of access to vast amounts of user interaction data, which can help train models to identify subtle cues. However, the cat-and-mouse game between generators and detectors is likely to intensify.
Industry perspectives and future outlook
Industry analysts have praised LinkedIn for taking a stand, but caution that the devil is in the details. Without clear metrics on false positives and user appeal mechanisms, the system could cause unintended harm. Some experts suggest that LinkedIn should implement a feedback loop where users can contest demotion of their posts, similar to how other platforms handle content appeals.
The move also reflects a growing recognition that AI-generated content, while useful in many contexts, can degrade the quality of online discourse if left unchecked. As generative AI becomes more accessible, the responsibility falls on platforms to maintain authenticity. LinkedIn's decision to suppress rather than remove may strike a balance between freedom of expression and quality control, but it remains to be seen whether the 94 per cent accuracy claim holds up in real-world deployment.
LinkedIn has not shared how quickly the rollout will happen, noting only that it could take several months before users see less low-quality AI material in their feeds. The platform is taking a cautious approach, presumably to avoid disrupting the user experience or alienating content creators who rely on AI tools legitimately. The coming months will be a test of whether LinkedIn can successfully filter out the noise without silencing valuable voices.
Source: TNW | Apps News