How the LinkedIn Algorithm Works — and How to Make It Work for You

Most advice about the LinkedIn algorithm is astrology with a blue palette. 'Post Tuesday at 9am.' 'Use exactly three hashtags.' 'The algorithm hates external links.' Some of these contain a grain of truth; most are superstition repeated until they feel like fact. LinkedIn's own engineering and creator teams have published enough over the years to describe the actual shape of the system — and independent researchers have measured the rest. This is that picture, without the astrology.

How does a LinkedIn post actually get distributed?

LinkedIn has described its feed ranking publicly through its engineering blog and its creator guidance. Simplified, your post passes through three gates:

1. The quality filter. Within moments of posting, automated classifiers label your post as spam, low quality, or clear. Obvious engagement bait ('comment YES if you agree'), walls of hashtags, and recycled viral text get throttled here before a human ever sees them.
2. The test audience. Your post is shown to a small slice of your network — people who follow you, have interacted with you recently, or match the post's topic. This is the moment that decides everything: the algorithm is watching whether those people stop, read, and respond.
3. The expansion. If the test slice dwells and comments, the post graduates to second- and third-degree networks, hashtag and interest followers, and 'people like your engagers' lookalike surfaces. Distribution compounds in waves — which is why a post can be quiet for two hours and then take off.

The practical consequence: your first hour matters, but not because of a countdown timer. It matters because the test slice's behavior is the input to every later decision. A post that earns real comments from ten relevant people beats a post that collects a hundred reflexive likes.

What signals does the LinkedIn algorithm actually reward?

Across LinkedIn's own statements and the large independent feed studies (notably Richard van der Blom's annual algorithm research), the same handful of signals keep showing up at the top:

• Dwell time. LinkedIn confirmed years ago that it measures how long someone lingers on your post — it's the closest proxy for 'this was worth reading.' Long text posts that hold attention and documents that get swiped both feed this signal.
• Comments with substance. Replies over a few words, and especially threads where the author responds, weigh far more than likes. LinkedIn has said it ranks conversation as a primary outcome of a healthy feed.
• Relevance. The system matches your post to viewers by graph proximity, declared interests, and past behavior. A niche post shown to the right 500 people outperforms a generic post sprayed at 5,000.
• Consistency and identity. Accounts that post on a stable theme train the system on who to show them to. Accounts that lurch between topics reset that learning every week.
• Native behavior. Posts that keep people on LinkedIn align with what the feed is optimizing for; posts that shove people off-platform start with a headwind.

Notice what's not on the list: hashtag counts, posting times, follower totals, company-page budgets. Those are margin notes, not levers.

What does the algorithm punish?

• Engagement bait — explicitly named in LinkedIn's guidance. 'Like if you agree' posts get down-ranked, and repeated offenders get account-level throttling.
• Engagement pods. Coordinated groups produce unnaturally timed, off-topic engagement that classifiers recognize. The short-term bump costs you long-term trust in the system.
• Heavy external linking, especially in the post body of every post. This one is real but overstated: an occasional link in a post that earns dwell anyway is fine; a feed that is nothing but link drops will underperform.
• Rapid-fire posting. Multiple posts inside a short window cannibalize each other's test audiences. One strong post a day beats three thin ones.
• AI sludge. Generic, same-shaped content gets low dwell, and low dwell is indistinguishable from low quality to the system. The penalty isn't for using AI — it's for publishing what nobody stops to read.

So what should you actually do differently?

Translate the mechanics into operating rules and the algorithm stops being a mystery:

1. Write the first two lines like they cost money. The feed truncates after roughly three lines; dwell starts at 'see more.' If the hook doesn't earn the click, nothing else you wrote exists.
2. Post on one clear beat. The algorithm needs pattern to route you; pick the two or three topics you want to own and stay on them for months, not days.
3. End with a real question, not a prompt. Questions that experts want to answer produce the long comments the system weights heaviest.
4. Stay in your own comments for the first hour. Author replies double comment counts and signal an active conversation — the cheapest distribution lever there is.
5. Format for the pause. Short paragraphs, one idea per line of sight, numbers where you have them. Dwell is a design decision before it's a writing one.
6. Publish when your audience is awake, then stop worrying. Timing shifts the size of your test slice, not the verdict it reaches.

The algorithm is not the audience. It's a delivery system trained on the audience. Win the human pause and the machine follows.

This is the same pattern that shows up across our research — from our study of 1,167 B2B LinkedIn posts to our read on how OpenAI's employees post: the accounts that grow are the ones built for human attention first, and the feed rewards them for it. our study of 1,167 B2B LinkedIn posts · how OpenAI's employees post

Frequently asked questions

How does the LinkedIn algorithm decide what to show?

It ranks posts by predicted value to each viewer, using signals like dwell time, comments, the relationship between author and viewer, and topic relevance. New posts are tested on a small slice of the author's network first and only expand if that slice engages.

Did the LinkedIn algorithm change recently?

LinkedIn tunes its ranking continuously, and the company has publicly described a deliberate shift toward 'knowledge and advice' — prioritizing posts with substance from people with relevant expertise over viral generic content. The core mechanics (test slice, dwell, conversation) have been stable for years.

Do hashtags still matter on LinkedIn?

Marginally. A few relevant hashtags help topic classification, but they are a weak signal next to dwell time and comments. No credible study shows hashtags rescuing a post people don't read.

Do external links kill your reach?

Overstated. Posts that exist only to push a link underperform because they earn little dwell, not because links are banned. If a link post is genuinely worth stopping for, it can still travel — but never make links the backbone of your feed.

How often should you post on LinkedIn?

Three to five times a week is the sweet spot for most professionals: frequent enough to compound relevance, spaced enough that posts don't compete for the same test audience. Consistency over months beats intensity over weeks.

Sources

• LinkedIn Engineering: 'Leveraging dwell time to improve the feed' — LinkedIn's own explanation of dwell as a ranking signal (engineering.linkedin.com).
• LinkedIn Creator guidance on feed distribution and engagement bait (linkedin.com/help and LinkedIn's creator programs).
• Richard van der Blom, annual LinkedIn Algorithm Report — the largest independent feed study, based on analysis of over a million posts.
• Invisible Keyboard, The B2B Video Reality Check — our analysis of 1,167 organic posts from 57 B2B software companies.