How the LinkedIn Algorithm Works

LinkedIn rebuilt how it ranks content, and the change explains almost everything about why advocacy works and why company pages struggle. This article covers what the platform now measures, what that means for the content you put in front of your team, and the program decisions that follow from it.

Understand what changed

LinkedIn replaced its ranking system with a model that reads meaning rather than counting signals. It looks at what a post is actually about and whether the person posting it is credible on that subject.

The consequences are measurable and they are large:

• Average views per post fell by roughly half.
• Company page organic reach declined sharply, and a company page post now typically reaches a small single-digit percentage of its own followers.
• Generic corporate content largely disappeared from feeds.

None of that applies equally to employee profiles, which operate under materially different rules. That gap is the reason an advocacy program produces reach a brand channel cannot buy.

Know the three things it measures

Topical authority. Does the person's profile match what they post about? Posts from people with recognizable expertise in a subject are distributed several times further than the same post from an account with no pattern. This is why role-based content groups matter: when a recruiter posts recruiting content and a seller posts customer insight, both out-distribute either of them sharing the same company announcement. See Plan Your Content Group Strategy.

Engagement quality. How long people stop and read, whether they save the post, and whether the comments are substantive. A save counts for many times what a reaction does. Formatted posts with line breaks and white space hold attention noticeably longer than a wall of text, so formatting is a ranking input rather than a style preference.

Relationship and relevance. Content from people the reader interacts with, about subjects they engage with. One-size-fits-all corporate updates are actively suppressed.

Plan around the first 90 minutes

When a post publishes, LinkedIn shows it to a small test slice of the author's connections. Across roughly the first 60 to 90 minutes it measures dwell time, comment velocity and saves, then decides whether to push the post to a wider audience over the following two to three days. Posts that fail the test are buried.

This is the single most actionable fact for a program owner, because it means timing is a coordination problem you can solve. When several advocates share and comment substantively inside the same window, those posts clear the threshold. Marking content as important, so it triggers notifications, is how you create that window on purpose.

Curate the formats that travel

The format ranking changed, and most internal training decks are still on the old one.

• Documents and carousels are now the strongest format by a wide margin, because swiping generates the dwell time the system rewards most. If you have a case study, framework or one-pager, load it. See Share a Document to LinkedIn.
• Text posts earn the highest ratio of comments to impressions, which makes them the right tool when you want conversation.
• Image posts perform reliably, roughly double plain text.
• Native video has declined and is no longer the default winner. It still works short, captioned and face to camera.
• Link posts reach the fewest people, because the platform has no interest in sending readers away.

The practical implication for content curation: a group stocked entirely with link posts will underperform no matter how good the articles are. Mix in documents and text prompts.

Know what actively suppresses reach

• Plain reposts. A reshare with nothing added carries no new information, barely circulates, and lowers the sharer's future distribution. This is the single most important thing to train out. See Why a Plain Repost Does Not Work.
• Hashtag stuffing. Hashtags now behave as search keywords rather than a discovery mechanism. Three to five specific ones. Ten or more can trip spam filtering.
• Content that has not been through a human. The platform detects unedited machine-generated text. AI Share Copy is a drafting tool, and the last edit should be the advocate's.
• Coordinated liking. Engagement pods are detected and discounted. Coordinated commenting by people who actually read the post is a different thing and it works.
• Bursts followed by silence. Cadence is part of what earns distribution.

Turn it into program decisions

1. Set the cadence expectation at two to three posts a week, not as many as possible. See How Often to Post, and When.
2. Make commentary non-optional in your training. Pre-written share copy exists to make it easy, not to be posted verbatim.
3. Organize content by role, so people post inside their subject.
4. Create shared posting windows for the content that matters most.
5. Report on engagement and clicks rather than impressions, which are estimated. See Impression Analytics.

FAQ

Why did our reach drop even though sharing went up?

Almost always a mix of two things: the platform-wide decline in views per post, and a high proportion of shares going out with no added commentary. The second is the one you can fix.

Do hashtags still help?

As search keywords, yes. As a way to reach people who do not already follow the author, no. Three to five specific ones, chosen so the right reader can find the post.

Should we still post from the company page?

Yes, for the record and for people who look you up. Do not expect it to carry distribution. That is what the program is for.

Is video still worth producing?

Short, captioned, face to camera, yes. Long uncaptioned video is the weakest use of production budget available.

What should we tell employees to do differently?

Three things: add your own commentary, stay recognizably on your subject, and comment on other people's posts. From Advocate to Employee Influencer is the article to send them.