LinkedIn details AI system for non-connection feed recommendations
LinkedIn’s new Profile Relevance model uses LLM-generated summaries to match posts with non-connections, improving feed metrics and giving creators a clearer route to relevant audiences beyond followers.
What happened
LinkedIn shared new information about how its feed ranks content from accounts users do not follow. In a post on the LinkedIn Engineering Blog, LinkedIn engineer Shiyong Lin explained the platform’s Profile Relevance model, which uses LLM-generated embeddings to match members and posts based on professional context.
The process summarizes each user’s profile into a prompt, taking into account each member’s profile information and posting history. It then builds a personalized guide of what a member is likely to be interested in. Lin said the system evaluates how well each post matches a member’s professional context across topic, location, job title, company, industry, seniority, and education.
“The system evaluates how well each post matches a member’s professional context across specific factors, including topic, location, job title, company, industry, seniority, and education.”
LinkedIn manually assessed these results before building a system that showed users more relevant posts from members they don’t follow in the app.
Key facts
- LinkedIn’s Profile Relevance model uses LLM-generated embeddings to match members and posts based on professional context.
- The system summarizes each member’s profile information and posting history into a prompt, then scores posts across topic, location, job title, company, industry, seniority, and education.
- Online experiments recorded a +0.54% lift in Feed Updates Viewed, a +0.94% increase in viral rate, and a -18.8% reduction in hide rate.
- For cold-start members with limited activity history, LinkedIn observed a +17.7% increase in viral engagement and a +1.8% lift in DAU.
- Lin said the system will benefit creators by ensuring their posts reach relevant audiences beyond their immediate network.
Our analysis
The most important shift for creators is that distribution is no longer just about follower count or existing network density. LinkedIn’s model explicitly scores content from non-connections against a member’s professional summary, which likely means a well-positioned post can travel to relevant users even if the creator does not have a large following.
The factors LinkedIn lists—topic, location, job title, company, industry, seniority, and education—suggest that LinkedIn is prioritizing professional specificity over broad engagement signals. A post that is clearly relevant to a defined audience segment likely has a better chance of being shown to that segment, while unclear or off-topic posts may be filtered out. The reported -18.8% reduction in hide rate also points to a less noisy feed; for brands and creators, this suggests that staying within a clear professional lane is more important than chasing volume.
The cold-start result is notable. A +17.7% increase in viral engagement for members with limited activity history suggests the model can evaluate content relevance without needing a long behavioral trail. That is a meaningful change for new accounts or new campaign pages, which have historically likely depended on building a connection graph before earning meaningful feed distribution.
What it means for operators
- Complete and maintain profile fields. Because the system summarizes profile information and scores against job title, company, industry, seniority, location, and education, incomplete profiles likely weaken the match between your content and the right audience.
- Pick a clear professional niche and stay on it. The model matches posts to a member’s professional context, so specific, topic-relevant content is likely to reach non-connections more reliably than broad or mixed-content posting.
- Reduce high-hide-risk content. The platform highlighted a -18.8% reduction in hide rate as part of a less noisy experience. Off-topic, clickbait, or overly promotional posts may be hidden by users and likely limit future relevance distribution.
- Use cold-start-friendly formats for new pages. LinkedIn observed a +17.7% increase in viral engagement and a +1.8% lift in DAU for members with limited activity history, so a new page should lead with highly relevant, clearly professional posts rather than waiting to build followers first.
Source:LinkedIn explains its improved content recommendation system — Social Media Today(2026-11-09)
Editor's note: prepared by the NoobClaw newsroom with AI assistance from the public report above. Facts are as reported by the source; the analysis is our view. Spotted an error? Contact us and we will correct it.