LinkedIn algorithm

LinkedIn does not publish its ranking weights, so anything stated as a precise percentage is a guess. What LinkedIn has said publicly, and what its engineering blog posts describe, is a two-stage system: a retrieval step pulls a candidate set of posts from your network and topics you engage with, then a ranking model scores those candidates on predicted response — will this person stop, read, react, comment, or click away. The feed you see is the top of that scored list, not a chronological record of who posted.

The practical consequence is that the first hour matters more than the first day. Early signals from people who see the post first — mostly your closest connections and recent engagers — feed back into whether the post gets shown to a wider set. This is also why comments tend to outperform reactions: a comment takes longer to write, implies the reader stayed on the post, and creates a second surface that can pull other people in. Dwell time and comments are correlated, and both are cheap for the model to measure.

The common misconception is that there is a single trick — a magic posting time, a banned word list, an outbound-link penalty you can beat by putting the URL in a comment. Those rules of thumb shift constantly and were never documented. The durable version is duller: post about a narrow topic consistently so the retrieval step knows who to show you to, write openings that earn the next line, and reply to your own comments. Everything else is noise dressed as strategy.

Related terms

← Back to the full glossary