LinkedIn: Slop Views Fell 40 Percent as Detection Expands

As part of ongoing efforts to fight AI slop on the platform, LinkedIn has expanded its AI slop classifiers to cover all posts distributed beyond a member's immediate network, according to a recent post from LinkedIn's Tim Jurka. The company says the classifiers detect AI slop, its term for generic and repetitive content, with 94% precision.

The detection system uses a teacher-student setup. Larger language models track new AI slop patterns and generate training data for smaller models that run quickly at scale. LinkedIn uses Low-Rank Adaptation (LoRA) to adjust a base model with task-specific adapters rather than build and serve a new model for each change. Jurka wrote that the approach lets the company respond as content patterns shift while using computing resources more efficiently.

LinkedIn also runs agentic workflows on a shared engineering platform. Each agent follows a policy that sets the criteria for evaluating content, such as whether a post is promotional, celebrates an achievement, or is timely. When an agent cannot resolve an example under its policy, it routes the post to human reviewers. Their guidance is then used to update the policy and the models.

Citing LinkedIn's Hari Srinivasan, Jurka wrote that members saw 40% fewer views of content classified as AI slop as of August 2026 compared with a few weeks earlier. LinkedIn plans to add more ways for members to report AI slop. The platform has 1.4 billion members.

Full story: LinkedIn

More FNN coverage: 
LinkedIn Adds an "AI Slop" Reporting Button
LinkedIn: 5 CMOs at Cannes Lions Said the Real AI Challenge Is People

REGISTER NOW

The future economy isn't just coming—it's already changing the rules for how financial services firms attract, engage, and earn customer trust. Step into the Fall Summit in partnership with Nuveen and TIAA, an exclusive gathering of senior marketing and communications leaders in New York on November 5th.