Short answer: The LinkedIn algorithm uses AI models to predict which posts each member will find valuable, drawing on hundreds of signals from their profile, network and activity. It looks beyond likes to how long people spend on a post (dwell time), the quality of comments and the engagement a post earns in its first 24 hours, and LinkedIn says it does not use demographics such as age, race or gender as ranking signals.
LinkedIn does not publish its ranking code, so most advice about the algorithm is guesswork. This explainer sticks to what LinkedIn and its engineers have actually said, then turns it into practical steps.
How does the LinkedIn feed decide what you see?
Like most large feeds, LinkedIn works in two broad stages. First it retrieves a pool of candidate posts that might interest you – from your connections and the people and Pages you follow, plus recommended posts from beyond your network. Then it ranks that pool by predicting how valuable each post will be to you, and builds your feed from the top.
The models doing this have changed quickly. In August 2025, LinkedIn’s VP of Engineering Tim Jurka wrote that the company had more recently started using large language models to help understand the context of posts. In March 2026, LinkedIn announced a new feed system: retrieval based on large language models, which finds relevant posts from inside and outside your network, and a “Generative Recommender” ranking model that reads more than 1,000 of your past interactions in sequence. In May 2026, LinkedIn’s CTO Erran Berger described generative recommenders replacing separate specialist models for the feed, jobs, People You May Know and ads with systems that read a member’s activity across the whole platform over time, and shaping which candidates are considered before ranking even starts.
LinkedIn researchers have also published work on 360Brew, a 150-billion-parameter foundation model trained on LinkedIn data to handle more than 30 ranking and recommendation tasks. Treat it as a sign of direction rather than a description of the live feed: the paper describes a research model, not exactly what your feed runs on today.
Which signals does the LinkedIn algorithm use?
LinkedIn says its systems weigh “hundreds of signals”, and that what members click, save and engage with shapes the feed every day. Its Help Center groups them into three types: identity (your profile, such as location, workplace and skills), content (what a post is about, how recent it is and how often it is viewed and engaged with) and activity (what you react to, comment on or share, who you interact with and the content you spend the most time viewing). These are the signals LinkedIn has spoken about publicly:
| Signal | What LinkedIn has said |
|---|---|
| Relevance to you | Rankings draw on your profile, your network and your activity across LinkedIn |
| Dwell time | A model predicts the chance you will skip a post; posts likely to be skipped are scored lower |
| Early engagement | Engagement in the first 24 hours influences how widely a post is distributed |
| Comments | Comments are a positive signal, and comment ordering is now personalised to each reader |
| Automated comments | Comments posted by scripts or plug-ins without human review are demoted and may not reach beyond the commenter’s own network |
| Demographics | Age, race and gender are not used as ranking signals; they are only used in testing for fairness |
Why does dwell time matter so much on LinkedIn?
Most LinkedIn members read without reacting, so likes and comments alone give the system very little to learn from. LinkedIn’s engineering team explained that clicks are noisy and viral actions are rare, whereas time spent is available for almost every post. They record two kinds of dwell time: time on the post in the feed, measured once at least half of it is visible, and time spent after clicking through.
From that data LinkedIn built a model that predicts the probability you will skip an update – scroll past it in less than a threshold time – and reduces the score of posts likely to be skipped. In online tests this led to fewer skipped posts, more clicks and more time in the feed. The team also stressed that the aim is “time well spent” rather than simply more time spent.
The practical lesson: a post that makes people stop and read – a strong first line, a clear structure, a useful document or image – earns a signal even from people who never react.
How much do comments and early engagement matter?
A lot, but quality counts. In June 2026 LinkedIn’s VP of Marketing, Davang Shah, wrote that engagement signals in the first 24 hours influence how widely a post is distributed, and that high-quality comments – ones that add insights, questions or different perspectives – are a positive signal for both the LinkedIn algorithm and AI retrieval systems. LinkedIn’s own posting checklist advises engaging with comments within the first hour of posting.
At the same time, LinkedIn has moved against fake conversation. In February 2026 it said automated comments would be removed from the default “Most relevant” view, could be stopped from reaching beyond the commenter’s direct network, and that repeat offenders could face account restrictions. It has also said it is tackling engagement pods – groups that agree to like and comment on each other’s posts. Our explainer on engagement pods and paid engagement covers the trade-offs, and our guide to LinkedIn engagement has tactics for earning real comments.
Does LinkedIn penalise external links, hashtags or posting too often?
External links: LinkedIn has not published a rule that penalises posts with links, despite years of debate. If you want to test it, compare similar posts with the link in the post, in a comment and left out, and judge by your own analytics.
Hashtags: LinkedIn’s marketing team has advised against opening a post with hashtags, because LinkedIn uses the first line to generate the post’s URL and a hashtag there creates a generic one. Open with your keywords instead. Our article do hashtags still work looks at the bigger picture.
Frequency: LinkedIn’s own checklist calls consistency “one of the biggest drivers of reach” and suggests two to three posts a week. For other platforms, see how often to post on each platform.
Articles and newsletters: long-form formats live alongside regular posts. LinkedIn’s marketing team suggests 800 to 1,200 words for Articles and 200 to 300 words for posts, with roughly one article and two to three posts a week.
How do you work with the LinkedIn algorithm?
- Write for your network first. Relevance to the people who know you drives early engagement; our guide to LinkedIn followers vs connections explains who sees your posts.
- Earn the stop. Make the first two lines specific enough that people pause, then reward them with substance so dwell time follows.
- Invite real discussion and reply to comments within the first hours.
- Stay consistent rather than posting in bursts.
- Avoid shortcuts LinkedIn targets: automated comment tools, pods and engagement bait.
If you want an established-looking presence while you build that habit, StreamElevate’s LinkedIn services only need a public profile, Page or post link – but no number of likes replaces posts your network genuinely wants to read.
Frequently asked questions
Is the LinkedIn algorithm chronological?
Not by default. The feed shows Top posts, ranked by predicted relevance, so you can see older posts above newer ones when LinkedIn expects them to be more useful to you. On desktop you can switch to Recent with the Sort by menu above your feed, or make Most recent posts your permanent view in Settings, to see posts from people you follow in reverse chronological order. LinkedIn has also tested alternative feeds, such as a Network tab.
Does LinkedIn show posts to people outside my network?
Yes. Retrieval includes recommended posts from beyond your connections and follows, and LinkedIn says its newer models widen that candidate pool.
Why did my LinkedIn reach drop?
LinkedIn reported that in 2025 posting rose 15% and comments 24% year on year, which it described as more competition in the feed. Combined with model changes and stricter handling of automated engagement, many accounts saw reach shift.
Does LinkedIn use AI to rank posts?
Yes. LinkedIn has said it uses large language models to help surface content, and its CTO has described a shift to generative recommenders and large-scale sequence models.
Sources: LinkedIn Engineering: Understanding feed dwell time, TechCrunch: what’s going on with LinkedIn’s algorithm, LinkedIn Engineering: Engineering the next generation of LinkedIn’s Feed, LinkedIn Pressroom: How LinkedIn is improving the Feed, Erran Berger (LinkedIn CTO): The next era of recommendations on LinkedIn, Tim Jurka (LinkedIn): How does the LinkedIn Feed work?, LinkedIn Help: LinkedIn relevance, LinkedIn Help: Sort your feed by Top or Recent posts, LinkedIn Marketing Blog: How to maximize AI visibility for your LinkedIn posts, LinkedIn for Marketing: posting checklist, Gyanda Sachdeva (LinkedIn): automated comments, Rishi Jobanputra (LinkedIn): Making LinkedIn conversations more relevant, arXiv: 360Brew