Social Rocket
LinkedIn has extended its detection of generic AI posts to content recommended beyond the immediate contact network. Engineering Vice President Tim Jurka cites a precision of 94 percent for this, dated 8 October 2026. For B2B marketers, the key question is which content will consequently reach new audiences less often – and what the figure does not prove.
94 percent precision is not infallibility
LinkedIn uses large language models as "teachers" that capture new patterns of repetitive content. Smaller models take over the review at scale. In addition, specialised AI agents work with content rules; unclear cases can be escalated to human reviewers. According to Jurka, the models are updated using, among other methods, efficiently adaptable LoRA techniques.
The 94 percent is a precision figure reported by the company for detecting "AI slop" in content distributed beyond one's own network. Precision describes the share of correct hits among the cases flagged as problematic. It does not mean that 94 percent of all AI posts are detected; that would require figures on the recall rate. LinkedIn also does not publish independent audits or error rates for specific languages.
What matters is substance, not the AI tool
LinkedIn defines "AI slop" as generic, repetitive or low-substance posts without a distinct perspective. According to the official help centre, AI-assisted wording is expressly permitted as long as experience, opinion and professional value remain recognisable. Low-value content is more likely to have its distribution restricted than to be deleted outright.
Back in September, LinkedIn reported 40 percent fewer views for content the platform itself classifies as "AI slop". More than one million members used the new feedback feature within the first two weeks. These are LinkedIn figures, not a controlled study of the success of the individual report button. The difference from the already documented approach against automated engagement and fake interactions is important: this time the platform explains in more concrete terms how it technically assesses content quality.
Ads also receive slop feedback – with different consequences
On LinkedIn, users can also select "Seems like AI slop" for feed ads. According to LinkedIn Marketing Solutions, this currently does not apply to the right-hand sidebar or the Audience Network. For ads, LinkedIn currently uses this feedback only for internal product improvements. Advertisers receive neither a notification nor a corresponding metric in reporting. The button should therefore not be taken to mean that individual campaigns are automatically devalued.
What B2B teams should check now
The practical risks lie less in using AI than in interchangeable text patterns: interchangeable promises of success, automated comment series and posts without original observations. A useful check before posting: does it contain a verifiable figure, a documented example, a professional decision or an independent perspective?
For performance analysis, it is also worth looking at impressions from one's own network and beyond separately, insofar as the available analytics allow this. LinkedIn's planned network feed and the shift from the AI rewriter to a proofreader fit the same quality strategy. In the B2B buying process, credible expert voices may matter more than high posting volume.
Sources: Tim Jurka / LinkedIn, 8.10.2026; LinkedIn News, 15.9.2026; LinkedIn Help for advertisers; Social Media Today, 8.10.2026.
Image: official title image for the engineering article, LinkedIn / Tim Jurka.