Nobody outside LinkedIn has read the algorithm's source code, and the people who have are not allowed to tell you. So every "the LinkedIn algorithm rewards X" claim you have ever read is either a leak, a guess, or a guru's hunch dressed as fact.

We took a different route. We cannot read the code, but we can measure its outcomes at scale. We looked at 1,207,853 LinkedIn posts published over the last 12 months (reshares and deleted posts excluded), and asked a simple question of each ingredient a post can have: when this thing is present, does the post earn more or less? Length, format, links, hashtags, hooks, whether it reads like AI wrote it, polls, and the size of the account behind it. The answers are remarkably consistent, and several of them contradict advice you have heard a thousand times.

One honest caveat first, because it governs everything below: these are correlations measured at scale, not controlled experiments. A post is not a coin flip. People who write 400-word posts also tend to be more invested writers; that is a confounder we name every time it matters, and where we can, we re-run the numbers inside follower bands to check the effect survives. Read this as "what wins on LinkedIn, measured," not "what LinkedIn's ranking function literally does."

Here are the levers, ranked by how hard they pull: