LinkedIn Algorithm 2026: What LinkedIn Actually Confirmed About Reach, Dwell Time and Expertise

No platform attracts more confident advice than LinkedIn. Post in the first hour, never use links, comment on ten posts before publishing, always use carousels.

Very little of it traces back to LinkedIn. This guide separates what LinkedIn has actually published about its feed from what has been invented around it, and explains what that means for the posts you write.

What LinkedIn confirmed in 2026

On 12 March 2026, LinkedIn’s engineering blog published the most detailed account it has given of how the feed works. Written by Hristo Danchev, it described replacing several separate systems with a new design:

• One retrieval system, using embeddings generated by large language models, to find the posts each member might want to see.
• A new ranking model, which LinkedIn calls a Generative Recommender, that looks at more than a thousand of a member’s past interactions rather than judging each post in isolation.
• Fast-updating signals, so that a post’s popularity and freshness are refreshed within minutes.

A separate LinkedIn post the same month focused on authentic, relevant conversations, and LinkedIn has said it can detect coordinated engagement and low-effort engagement bait.

Two stages: being found, then being ranked

Stage	What it decides	What it means for you
Retrieval	Whether your post is even a candidate for a given member	Clear topics and a profile that matches them help the system understand who your post is for
Ranking	Where your post appears among the candidates	The member’s own history decides most of this, so relevance to the reader beats broad appeal

The practical shift is from audience size to audience fit. A post that is clearly about one subject, from someone visibly experienced in it, has a better chance of reaching the people who care, whatever the size of the author’s network.

Why posts can resurface days later

Because signals are refreshed continuously, ranking is not a one-off decision made in the first hour. A post that starts slowly can be shown to more people later as readers spend time on it and conversations build.

The first hour still matters, but as a starting point rather than a verdict. Deleting a slow post and reposting it throws away whatever it had begun to earn, and repeated reposting can look like duplication.

Dwell time

LinkedIn’s engineers wrote about dwell time as a ranking signal in 2020, and it remains part of the picture. They described two kinds:

• Dwell on the feed, which starts counting once at least half of a post is visible as someone scrolls.
• Dwell after the click, the time spent on content once it is opened.

The idea is simple: a quick like is a noisy signal, while time spent reading suggests genuine interest. Someone who reads your post carefully without reacting is still telling LinkedIn something.

• Open with a specific claim or situation that makes the reader want the rest.
• Write in short, readable paragraphs rather than one dense block.
• Give something worth thinking about: a number, a decision, a lesson that took real experience to learn.

Expertise and your profile

LinkedIn’s published research describes using member profile information, such as headline, skills and experience, in how members and content are represented. LinkedIn has not published exactly how much this matters in ranking.

The safe reading is consistency. A profile that says you work in logistics and a stream of posts about logistics tell the same story. Constantly changing subjects makes it harder for any system, and any reader, to know what you are worth following for. The same principle drives how Google judges expertise, covered in the E-E-A-T scoring rubric.

Comments and conversations

• Substantive comments that add a view or a question are what LinkedIn says its feed is designed to encourage.
• Replying to commenters turns a post into a conversation, which keeps it relevant to the people taking part.
• Engagement pods, where groups agree to like and comment on each other’s posts regardless of interest, are something LinkedIn says it detects.
• AI-written comments add little to a conversation and are a known target.

Profiles, company pages and newsletters

• Personal profiles carry the expertise signals described above, which is why individual experts often reach further than the companies they work for.
• Company pages tend to see lower organic reach in third-party studies. LinkedIn has not published comparative figures, but pages rarely have a single, consistent area of expertise.
• Newsletters notify subscribers when a new edition is published, which gives them a route to readers that does not depend entirely on the feed.

For most businesses, the strongest approach combines them: people sharing what they know from their own profiles, and the company page supporting rather than replacing them.

What is not confirmed

• “360Brew runs the feed.” 360Brew is the name of a LinkedIn research model. LinkedIn’s March 2026 engineering post does not say it runs the feed.
• Exact signal weights, such as dwell time being worth a precise multiple of likes. LinkedIn has not published any.
• Fixed feed percentages, such as a set share from your connections. These come from third-party studies, not LinkedIn.
• “Links kill reach.” LinkedIn has not announced a penalty for links; results vary by audience, so test for your own.

When a statistic about LinkedIn has no source, treat it as a guess. Some are reasonable guesses based on real data; none should be planned around as fact.

A worked example

Imagine two consultants posting about pricing on the same day.

• The first posts “7 pricing tips every founder needs”, a list that could have been written by anyone, with a request to comment for a free template.
• The second describes raising a client’s prices by 30 percent, what the client feared, what actually happened to sales, and the one question that changed their mind.

The first may collect quick reactions, but readers scroll past the list and the comment request looks like bait. The second holds attention, invites real replies and matches the consultant’s stated expertise. It is the kind of post the rebuilt feed is designed to show to the right people.

A weekly routine that fits the new feed

1. Two or three posts a week on the subjects your profile says you know.
2. Reply to comments the same day, so conversations keep going.
3. Comment thoughtfully on others’ posts in your field, adding a view rather than praise.
4. Review analytics monthly by topic, and do more of what your target audience actually reads.

Consistency on a narrow set of subjects does more than volume. It gives the system, and your readers, a clear picture of why your posts are worth seeing.

Diagnosing a drop in reach

1. Compare engagement rate, not just impressions. Fewer views with more responses can mean better targeting, not a problem.
2. Check who is viewing in your post analytics. If the audience shifted away from your field, your topics may have drifted.
3. Compare topics with your profile. Posts far from your stated expertise may reach fewer relevant people.
4. Stop any coordinated engagement. If you relied on pods, the reach they once produced is unlikely to return.
5. Look at what changed: format, length, posting rhythm, or the subjects you write about.

Other platforms reward the same underlying thing, attention earned rather than borrowed. See the TikTok algorithm in 2026, the YouTube algorithm guide , the Instagram algorithm guide , the Facebook algorithm in 2026 and the X algorithm in 2026.