LinkedIn is launching a new AI-powered feed ranking system that uses large language models and GPUs to analyze post content and surface more relevant updates to its 1.3 billion members.
Why we care. Understanding how LinkedIn surfaces content is critical if you want your posts — or your brand’s — to be discovered. The new system prioritizes topical relevance and engagement patterns, LinkedIn said. Posts that demonstrate expertise and align with emerging professional conversations may travel farther across the network — even without existing connections.
The details. LinkedIn rebuilt much of its feed recommendation system using large language models, transformer models, and GPU infrastructure. The overhaul centers on two systems: retrieving relevant posts and ranking them in the feed.
Unified retrieval system. LinkedIn replaced several separate discovery systems with a single LLM-powered retrieval model.
Previously, feed candidates came from multiple sources, including network activity, trending posts, collaborative filtering, and topic-based systems.
The new approach uses LLM-generated embeddings to understand what posts are about and how they connect to your professional interests.
Now, LinkedIn can link related topics even when they use different terminology. For example, engagement with posts about small modular reactors could surface content about electrical grid infrastructure or renewable energy.
Ranking that follows your interests. After retrieval, LinkedIn ranks posts using a transformer-based sequential model. Instead of evaluating posts independently, the model analyzes patterns across your past interactions — including likes, comments, dwell time, and other signals.
This helps LinkedIn detect how your professional interests evolve and recommend content that reflects those shifts.
System performance and infrastructure. The system runs on GPU infrastructure designed to process millions of posts while keeping feeds fresh.
The architecture can update content embeddings within minutes and retrieve candidates in under 50 milliseconds, LinkedIn said.
Improving feed quality and authenticity. LinkedIn also announced updates to improve content quality:
Cracking down on automated engagement. LinkedIn is taking action against comment automation tools, browser extensions, and engagement pods that create inauthentic conversations. These tools violate platform rules and undermine real professional discussions, LinkedIn said.
Reducing engagement bait and generic posts. LinkedIn plans to show less content designed purely to drive comments or clicks — including posts asking people to comment “Yes” to boost reach, posts pairing unrelated videos with text to game distribution, and recycled thought-leadership with little substance.
Helping new members personalize their feeds faster. LinkedIn is testing an “Interest Picker” during signup that lets new users choose topics such as leadership, job search skills, or career growth, helping deliver relevant content from day one.