Feed

The Feed's content comes from the formats LinkedIn supports and the relationships a member has built, and LinkedIn Ireland's reports list the formats as text posts, images, recorded video and livestreams, with ads and suggestions for people, Pages, courses and jobs also appearing111. LinkedIn Help and a 2021 LinkedIn blog guide say the Feed is built from the people a member is connected to, the people and organizations they follow, and topics they have shown interest in312. A member automatically follows their 1st degree connections, meaning people they are directly connected to, and can also follow someone without connecting, which puts that person's posts and articles on their homepage13. When a member follows a Page, all posts the Page shares appear in their homepage feed14. LinkedIn Help says its systems track social actions such as writing a post, liking and commenting and use them in its algorithms, and recommended items can carry labels such as Trending, Recommended for you or Popular3.

A LinkedIn research paper from February 2026 describes how a pool of possible posts becomes one ordered Feed6. First, candidate posts, meaning the posts considered for a member's Feed, are retrieved from multiple sources and combined6. A ranking model then predicts the likelihood of each action, such as dwelling on a post longer than a set time or liking, commenting on or sharing it, and the resulting scores are combined using weights in an objective function that is used to rank the posts6. Extra business logic is then applied to the ranked list6. LinkedIn Help says the content served is chosen using a member's previous interactions such as comments, reactions and time spent, their profile information such as job title, location and industry, and how likely they are to engage, and Tim Jurka adds that the models also use skills and experience a member chose to share and content the member has or has not reviewed45. Tim Jurka and LinkedIn Engineering wrote in March 2026 that traditional ranking models judged each impression, meaning each showing of a post to a member, on its own, while the Generative Recommender model reads more than a thousand of a member's past interactions as one sequence, so a run of related posts a member engaged with over several days counts as a continuing interest rather than separate events515.

Members can change what the ranked Feed shows them by hiding posts, unfollowing or following sources and choosing how the Feed is ordered78. The sorting options differ inside and outside the European Economic Area (EEA)8.

LinkedIn's leaders have described what the Feed should reduce through examples521. Tim Jurka wrote in March 2026 that the Feed will show less of "Comment 'Yes' if you agree" prompts, less of posts with a video unrelated to the text, which should no longer gain additional reach, and less of recycled thought leadership posts with little substance5. Oscar Rodriguez wrote in September 2025 that engagement pods are coordinated groups who like, comment and share posts to boost virality, that LinkedIn flags posts that look artificially boosted and limits their reach in the feed, and that it cracks down on browser extensions and plug-ins that enable it21. He added that when LinkedIn detects patterns suggesting a person or company is creating inauthentic engagement around their own posts, it can affect how that content is distributed21. Tim Jurka wrote in October 2026 that LinkedIn is evolving its systems to recognize and reduce repetitive, generic content often called "AI slop"24. Laura Lorenzetti wrote in March 2026 that over 150 editors and creator managers work inside the feed, spotting emerging topics and helping shape it alongside the AI2526.

The ranking described above runs on professional signals, and LinkedIn says it regularly audits the ranking model for equitable treatment across its member base1527. Tim Jurka wrote in August 2025 that the system factors in the author's professional identity, such as experience, profile content and skills, and none of the author's demographic information such as age, race or gender27. LinkedIn Engineering wrote in November 2025 that demographic data is used to test and measure the Feed before critical model launches28. One test, called creator allocation, checks whether posts of similar engagement and quality from one group are systematically ranked lower than another's, and the other checks whether Feed quality is systematically worse for one demographic, such as women seeing more irrelevant items than men28. Tim Jurka added that LinkedIn also checks that similarly engaging posts are ranked the same way across its member base27.