The LinkedIn algorithm in 2026, explained through the first 60 minutes of a post

You do not need LinkedIn's source code to understand how distribution works. You need to watch what happens in the hour after you publish, across enough posts to see the pattern. The pattern is consistent enough that creators plan around it, and it explains almost every "why did this post die?" question we get.

This is not a leaked spec. It is a working model built from LinkedIn's own engineering write-ups, from what the product visibly does, and from watching thousands of posts go out on schedule through ShipMyPosts. Where the model is inference, we say so.

Phase 1: the test audience (minutes 0–60)

When you publish, LinkedIn does not show the post to all of your connections. It shows it to a small initial group, mostly people who interacted with you recently, plus a sample of your connections and followers, plus a few people who engage with the topic. This is a test. The question the system is asking is simple: do people who see this stop, read, and respond?

Three things follow from this:

• Your recent engagers are your first audience. People who commented on you last week are the ones who see you first this week. That is why cadence compounds, see how often to post.
• The test is small. A few hundred impressions for a mid-sized account. Early numbers look tiny and then either climb or stop.
• The test is timed. If your audience is asleep during it, the signals are weak, and weak signals stop distribution. Hence morning posts, see the best time to post.

Phase 2: the signals, roughly in order of weight

Not all engagement is equal. Based on what LinkedIn has said about prioritising "meaningful" interactions and on what we observe, the ranking looks like this:

• Comments with substance. A reply longer than a few words, from an account with a real history, is the strongest positive signal. "Great post!" barely counts.
• Dwell time. How long people stay on the post: expanding "see more", reading to the end, swiping a carousel. LinkedIn has explicitly described dwell time as a ranking input. It is why walls of text die (people bounce) and why carousels travel (people swipe).
• Reposts with a comment. A repost that adds a sentence is a strong endorsement; a bare repost is weaker.
• Reactions. Positive but cheap. They matter most as a *speed* signal, early reactions tell the system the post is not being ignored, rather than as a quality signal.
• Saves. A quiet signal that the post has lasting value. Carousels and checklists earn them.
• Clicks on external links. Neutral to negative for reach. Sending people off-platform is not what the feed is optimised for, which is why "link in the first comment" became the norm.
• Negative signals. "Hide this post", "unfollow", and reports are strongly negative. Engagement-bait ("comment YES to get the PDF") has been targeted explicitly since 2024.

Phase 3: widening (hours 1–48)

If the test audience responds, the post is shown to a wider circle: more of your first-degree network, then second-degree connections of the people who engaged, then people who follow the topic or hashtag. Each circle is a new test, the post has to keep earning it.

This can continue for 24 to 48 hours. A post that keeps collecting comments keeps getting distributed; some posts get a second wind on day two when a well-connected person comments. If the first audience shrugs, the post quietly stops within a few hours and no amount of later activity revives it.

One more thing LinkedIn has said publicly: since 2023 it favours knowledge and advice from people with demonstrable expertise, shown to people for whom the content is relevant, over broad viral content. In practice: a post about hiring engineers travels further among people who hire engineers than among everyone. Niche is not a limitation; it is how distribution works now.

What dwell time actually measures

Dwell time is not "how long the post is". It is how long the reader chooses to stay. A 300-character post that makes people stop and think can have a higher dwell time than a 1,300-character post they skim. Three things raise it:

• A hook that earns the "see more" tap. The expansion itself is a dwell signal. See 25 hook formulas.
• A shape the reader can feel: story with a turn, claim with evidence, list with a spine.
• Short lines and white space, so the eye keeps moving down instead of bouncing off a wall.

Why edits and links hurt

Edits. Editing a post shortly after publishing appears to interrupt the evaluation; in our observation, posts edited in the first hour reach less than identical posts left alone. Proofread before publishing, not after. If you must fix a typo, do it after the first few hours or accept the cost.

Links. A post whose main action is "click this link" is telling LinkedIn that the destination is more valuable than the post. Put the link in the first comment, and make the post worth reading on its own.

Tagging. Tagging ten people who do not respond is a mild negative. LinkedIn has said unanswered tags can reduce distribution. Tag only people who will actually reply.

What boosts can and cannot do

Because the first hour is a test, giving a good post an early push is legitimate, as long as the engagement is real. Fake accounts and automated engagement pods get detected and penalised; LinkedIn has been explicit about this since 2024. Real reactions from real members, delivered in the first minutes, simply make sure the test audience is not silent.

That is the model behind the per-post boost in ShipMyPosts: real people, real reactions, in the window the algorithm watches. What a boost cannot do is rescue a post nobody wants to read. It raises the floor of the test; the post still has to earn the widening.

The first-hour playbook

Run this on every post you care about:

• Publish at a fixed time when your audience is awake: weekday morning in their timezone.
• Block the next thirty minutes. Do not publish and walk into a meeting.
• Reply to every comment, fast, with substance. A reply is a comment; a conversation is many comments. Ask the commenter a question back.
• Do not edit. Proofread before.
• Put the link in the first comment, if there is one.
• Send the post to three people who will genuinely reply: not "please like", but "you'll have an opinion on this". Real conversation, not a pod.
• Do not post again for the rest of the day. Your own posts compete with each other.

Reading your results

Judge a post at 24 hours, not at 10 minutes. Look at:

• Impressions vs your median. Above median: the test passed. Far below: the hook or the timing failed.
• Comments vs reactions ratio. A high ratio means the post provoked thought; those posts widen longest.
• Profile views and follows. The lagging indicator that a post converted strangers into readers.

And then stop reading results. One post is a sample of one. The pattern shows over twenty.

Three post autopsies

The model is easiest to see on real cases. Names removed, patterns kept.

• The one that died at 200 impressions. A good story about a hiring mistake, published at 9 pm, with the lesson in the first line and the story after. The test audience was asleep; the few who saw it got the lesson without the tension and scrolled. Same post, republished three weeks later at 8 am with the scene as the hook, reached twenty times as many people. Timing and hook, not content.
• The one that spiked and stopped. A contrarian take that earned 40 reactions in an hour and three comments. Reactions said "not ignored"; the lack of comments said "not worth discussing". Reach flattened by hour three. The fix was the ending: a version that closed with a pointed question earned 30 comments and kept widening into day two.
• The one that grew on day two. A carousel with a strong cover and a save-worthy checklist. Modest first hour, but saves and dwell time kept accumulating, and a well-connected reader commented the next morning. Second wind. Carousels are the format most likely to do this, see the carousel guide.

What changed recently, and what did not

LinkedIn's feed has shifted a few times in the last two years: a stated preference for knowledge and advice over viral content (2023), explicit action against engagement bait and "comment to receive" posts (2024), and steady de-emphasis of external links. What has not changed is the mechanism: a small test, early signals, staged widening. Strategies built on the mechanism, specific posts, fixed morning slot, thirty minutes in the comments, survive every update. Strategies built on a loophole do not.

The algorithm is not a black box. It is a room of people you get to talk to first. Say something worth answering.