The LinkedIn algorithm ranks each post by predicting how likely a given viewer is to find it relevant and worth engaging with, then it orders that viewer's feed by that prediction. The strongest observable inputs are early engagement in the first hour, dwell time (how long people stop and read, including whether they tap see more), and how closely the post matches the viewer's network and stated interests. LinkedIn has never published the full formula, so treat everything below as observed mechanics plus what its own engineering has stated, not a leaked recipe. I run my own account on these signals and draft against them daily. The honest headline: distribution is the game, and the fold is where most posts die.
LinkedIn does not fully document its ranking, so this is observable mechanics plus what its engineers have said. Reach is ranked per viewer by predicted relevance. Early engagement in the first hour, dwell time at the see-more fold, and network relevance are the load-bearing signals. Real comments beat vanity likes. What suppresses reach: engagement bait, pods, and generic likely-AI writing, which underperforms human writing per Originality.AI. Median impressions fell about 47% year over year per AuthoredUp. Write from real work so the substance survives the fold.
Start here: nobody has the full formula, including me
Any post that claims to reveal the exact LinkedIn ranking formula is selling you something. LinkedIn has never published its weights, and it changes them. What we do have is threefold: LinkedIn's own engineering blog and public statements, large third-party studies of millions of posts, and the pattern you see across your own account when you post the same way for months.
So read this as mechanism, not magic. I am describing the signals that are observable and repeatedly confirmed, plus the few things LinkedIn has said out loud. Where I am giving an operator's opinion rather than a documented fact, I say so. That honesty is the whole point. You cannot game a system you refuse to look at clearly.
The one-line model: relevance predicted per viewer
The useful mental model is a relevance prediction. For every post that could appear in your feed, LinkedIn estimates how likely you specifically are to find it worthwhile, then orders your feed by those estimates. There is no single global score for a post. The same post can rank high for one person and never surface for another.
That prediction leans on a few things the platform can measure well: signals from the people who see the post first, how long viewers spend with it, and how close the author and topic sit to your own network and interests. When a post clears the early bar, LinkedIn shows it to a wider slice; when it stalls, distribution quietly stops. That test-and-widen loop is the mechanism most creators actually feel, even if they never see the numbers behind it.
The golden hour: early engagement is a test, not a reward
When you publish, LinkedIn shows the post to a small initial audience, largely people already connected to you or in your immediate network. What that group does in the first stretch, often described as the first hour, heavily shapes how far the post travels next. Fast, real engagement reads as a signal to widen distribution. A flat opening reads as a reason to stop.
This is why the first hour matters more than the total. A post that gets ten thoughtful comments in forty minutes will usually outrun a post that collects the same ten spread over two days. My launch post reached under 200 people in its first hour, but roughly one in eight of them engaged, and that rate is what kept it moving. Low reach with a strong early rate is a healthy signal, not a failure.
The practical read: publish when your actual audience is awake and near their phones, then be present. Answering the first comments yourself, quickly and with substance, is one of the few levers you fully control.
Dwell time and the see-more fold: your real character limit
Dwell time is how long someone stops on your post instead of scrolling past. LinkedIn has publicly pointed to dwell time as a meaningful signal, and it makes sense: a like takes a quarter second, but stopping to read is a costlier, more honest vote of interest.
The single most important piece of real estate on a LinkedIn post is the see-more fold. The feed truncates long posts after the first couple of lines and hides the rest behind a see more tap. That tap is a dwell event. If your opening lines do not earn it, the rest of your post might as well not exist. The fold is the real character limit, not the 3,000-character maximum.
So the first two lines are not a warm-up. They are the whole pitch for the click. Lead with the specific, surprising, or useful thing. Bury the throat-clearing, or cut it. I check every draft against the fold before it ships, because that one line decides whether the substance underneath ever gets read.
You can see exactly where your post cuts off with a post preview tool, and keep your opening tight with a character counter. More on the mechanic in the dwell time glossary entry.
Relevance: why the same post reaches different people
LinkedIn is not trying to show every post to everyone. It is trying to show each person a feed they will find relevant, which means proximity and topic fit do a lot of work. Posts from people you interact with, in industries and on subjects you already engage, get a head start. Posts far from your graph face a much higher bar.
This has two consequences worth internalizing. First, a tight, consistent topic helps the system learn who you are for, which helps it find the right next readers. If you post about six unrelated things, you make the relevance prediction harder every time. Second, engagement from relevant people is worth more than engagement from anyone. Ten comments from people in your field signal more than a hundred likes from a giveaway.
You do not control the graph, but you influence it. Showing up in the right conversations, consistently, teaches the platform where you belong. That is slower than a hack and far more durable.
Comment quality beats vanity likes
Not all engagement is weighted the same. A like is the cheapest signal on the platform. A comment, especially a substantive one that itself gets replies, is far more expensive to produce and reads as much stronger interest. Comments also start threads, and threads generate more dwell time and more surface area for the post to be seen.
The honest version of this is about substance, not volume. A one-word great post comment is closer to a like than to a conversation. What moves a post is a real exchange: someone adds a point, you answer, a third person weighs in. That is the behavior the system is trying to find, because it is the behavior that makes the feed worth opening.
This is also why writing to provoke a genuine reaction outperforms writing to be agreeable. Take a clear position people can respond to. A post that starts eight real conversations will beat a post that collects eighty drive-by likes, every time.
What suppresses reach: bait, pods, links, and generic AI
Some behaviors reliably cost you distribution, either because LinkedIn has said it acts on them or because the pattern is clear across large datasets.
Engagement bait is the clearest case. LinkedIn has publicly stated it demotes posts that explicitly ask for likes, comments, or follows to manufacture engagement (comment YES below, tag three friends). The system is built to reward interest it did not have to beg for.
Engagement pods, groups that agree to like and comment on each other's posts on cue, are a bet against the relevance model. The whole system is trying to measure genuine interest from relevant people. Coordinated, off-topic engagement is exactly the noise it is designed to discount, and it can misfire by teaching the algorithm to show your work to the wrong audience.
External links are the most argued-about item, so here is the honest state of it. It is widely observed that posts driving traffic off-platform tend to reach fewer people than native posts, and LinkedIn has an obvious incentive to keep attention on LinkedIn. But this is a strong observed tendency, not a documented penalty with a published number. A common, low-risk practice is to put the link in the first comment rather than the body. Treat that as a sensible hedge, not gospel.
And then there is the content itself. Originality.AI's 2025 study of 3,368 posts from 99 influential profiles found that detectably-AI posts underperform human writing in most professional sectors. In a feed increasingly full of generic AI output, sounding like a specific human is a distribution advantage, not just a taste preference.
| What helps | What hurts |
|---|---|
| A first line that earns the see-more tap | Burying the point below the fold behind throat-clearing |
| Real engagement in the first hour from relevant people | A flat opening hour that signals the post to stall |
| Substantive comments and back-and-forth threads | Vanity likes and one-word replies with no conversation |
| A consistent, learnable topic and cadence | Six unrelated topics that muddy the relevance prediction |
| Specific numbers, real stories, a defensible position | Generic likely-AI writing, which underperforms per Originality.AI |
| Native posts that keep readers on-platform | Off-platform links in the body (observed to reach less, not a published penalty) |
| Engagement you earned by being interesting | Engagement bait (comment YES) and pods, which LinkedIn demotes or discounts |
What helps vs what hurts your LinkedIn reach in 2026 (observed mechanics plus LinkedIn's own statements)
The elephant: reach is down for almost everyone
Before you blame your own posts, know the baseline moved. AuthoredUp's analysis of more than 3 million posts found median impressions fell about 47% year over year, from 1,211 per post in June 2024 to 636 in May 2025, with the large majority of tracked users seeing declines. That is a platform-wide recalibration, not a personal penalty.
Part of it is simple dilution. When content supply surges, much of it likely-AI, and attention does not, every post's share of the feed shrinks. So calibrate your expectations to 2026 numbers, not to your best month two years ago. Judge yourself on engagement rate and conversations started, not on raw impressions you are unlikely to see again at the old scale.
I wrote a fuller diagnostic on the drop, including the four checks I run before blaming the algorithm, in a separate piece. The short version: if 98% of people share a symptom, the cause is the market, not you.
How to actually work with it: distribution is the game
Put the mechanics together and the strategy is boring in the best way. Write your first line to win the fold. Publish when your people are around, and be there for the first hour. Hold a cadence you can sustain, because consistency teaches the relevance model who you are for. Prefer starting conversations over collecting likes. Skip the bait and the pods.
The deeper move is upstream of all of it: substance. The reason generic content loses is that it has nothing only you could say. My best material comes from real work I actually did, and it survives the fold because there is a specific claim in the first line worth reading. If you write from your own projects, numbers, and arguments, you are automatically on the right side of the AI-content split.
That is exactly how I run my own account, and the tooling I built for it became Liftli. It works inside the AI you already use, drafts posts and comments from your voice notes and calls in your own extracted voice, and nothing ships without your one-tap approval. The algorithm rewards a specific human who shows up consistently. The job is to make that sustainable.
If you want the vocabulary, the LinkedIn algorithm glossary entry keeps the definitions current as the system changes.