Skip to content

BlogTrends for 20277 min read

LinkedIn algorithm 2026: the changes, explained

LinkedIn rebuilt its feed on language models in March. What it measures now, and how to write for it.

A VBS colleague writing a LinkedIn post on his laptop in a lounge chair

A feed that reads what you write

The LinkedIn algorithm changed fundamentally in 2026. On 12 March, LinkedIn's engineers described how they rebuilt the feed: a language model retrieves posts by meaning, and a second model ranks them on more than a thousand of the reader's past interactions. What someone read, where they lingered and what they skipped now weighs more than whom they follow.

For anyone using LinkedIn to win business, especially across borders, that changes how you write. Below: what LinkedIn itself published, the signals that count, what it means for different kinds of businesses, and how to write a post that makes it into the feed of the buyers you want.

  1. Your post

    Text, image, who shares it

  2. Retrieval by meaning

    A language model finds posts that match a reader's interests

  3. Ranking on behaviour

    More than 1,000 of the reader's past interactions

  4. Signals counted

    Long reads, comments, shares, clicks, skips

  5. The feed

    Relevant and fresh, never on age or gender

New since 12 March 2026: steps 2 and 3 run on language models instead of separate rules.

What changed on 12 March 2026

LinkedIn's engineering blog described the new feed in detail. Two things are new. First, a language model searches millions of posts for candidates that match a reader, based on meaning rather than keywords or network alone. Then a generative model ranks those candidates by looking at the sequence of more than a thousand of the reader's past interactions: which posts they read, where they commented, what they skipped. The hour you post mostly decides how fast those first readers arrive; the strongest windows for B2B are in the best time to post on LinkedIn.

SignalWhat LinkedIn measuresWhat it means for your post
ProfileThe reader's role, industry, skillsWrite for a clear reader, not for everyone
Engagement historyWhat the reader read and reacted to beforeStay with your topics so you return to the same readers
Long dwellsHow long someone stays with a postA post that holds attention beats one that only grabs it
Comments and sharesWhether someone adds to it or passes it onAsk a real question, share something useful
Clicks and skipsWhether someone clicks through or scrolls pastA first line that gives the answer
Relevance and freshnessWhether it fits the reader right nowPost when the topic is live

LinkedIn states explicitly that the feed never uses demographic attributes. It does not measure who you are; it measures what you read.

Why dwell time weighs more

That reading time counts is not new. LinkedIn explained in 2020 that it measures how long someone stays with a post and how likely they are to skip it. What is new is that the model now learns from the full sequence of what a person did, not from separate signals. A post that ten relevant people read carefully weighs more for those ten than a post with a hundred quick likes from people outside your topic.

A post people actually read

  1. A topic your reader struggles withA client question, a mistake you see often, a number from your own work.
  2. A first line that stops the scrollThe answer or the problem straight away, no warm-up.
  3. Text that holds attentionShort paragraphs, an example, a concrete figure; 200 to 300 words is enough.
  4. A question that invites repliesA real question to your reader, not a token 'thoughts?'.

What it means for your business

ReaderWhat changesWhat you do
Company outside Belgium targeting Belgian or EU buyersReach follows topic and languagePosts in the buyer's language; Dutch or French for Belgian decision-makers, English for international roles
Small business or service companyReach follows topics, not your networkThree fixed themes, posts from the founder and from the company page
Online storeProduct posts score lower than explanationsPosts on how your product solves a problem, behind the scenes
AgencyEngagement pods and tricks work less wellAdvise clients on substance and reading time, report on enquiries

What LinkedIn itself recommends

On 11 August 2026, LinkedIn's marketing blog gave B2B companies concrete guidance: posts of 200 to 300 words, articles of 800 to 1,200 words, two to three posts a week, and keep that rhythm for at least thirty days before drawing conclusions. Pick three to four core topics. That fits what the new feed rewards: recognisable topics and text long enough to be read. Plan the rhythm in a content calendar template.

Personal profile or company page

Posts from people usually reach more than posts from pages, because people respond to people. But a company page remains necessary: for ads, for jobs, and since 23 July 2026 for collaborative posts with up to five collaborators, people and pages together. A post by your founder and your page then appears to both audiences. For a company entering Belgium, a collaborative post with a Belgian partner is a direct route to their network. How to set the page up is in how to create a LinkedIn company page.

Questions people search about the feed

Did the LinkedIn algorithm change in 2026?

Yes. On 12 March 2026 LinkedIn described a new feed that retrieves posts by meaning with language models and ranks them on the reader's behaviour.

Do external links hurt reach?

LinkedIn has published nothing official on that. What it does measure is how long people stay with a post, and a post that only points elsewhere holds them for less time.

How often should a company post on LinkedIn?

In August 2026 LinkedIn advised two to three posts a week, kept up for at least thirty days, on three to four core topics.

Does LinkedIn rank on age, gender or location?

LinkedIn says the feed never uses demographic attributes.

Five myths you can drop

  • "Post on Tuesday at 8am."LinkedIn ranks on relevance and freshness for each reader, not on a fixed slot. Post when you can reply in the first hour.
  • "Engagement pods help."Quick likes from people outside your topic teach the model your post belongs with them, not with your buyers.
  • "Hashtags are the key."The new feed retrieves by meaning. A clear text on your topic says more than five hashtags.
  • "Carousels always win."A format wins when people stay with it. Ten thin slides do not.
  • "You need to post daily."LinkedIn itself suggests two to three times a week.

An example from our own work

For Business Centrum, a payroll office in Ghent, we did not run a LinkedIn campaign; we built a site with hundreds of guides on questions employers actually ask. In five weeks, its Google impressions went from 220 to 10,200 a day. The principle carries over to LinkedIn: whoever answers what their reader is looking for, concretely, gets shown more. See the Business Centrum case.

Before you hit post

  • The topic belongs to one of your three core themes.
  • The first line gives the answer or the problem.
  • There is an example or a number from your own work.
  • It reads well on a phone, in short paragraphs.
  • You are around in the first hour to reply.

LinkedIn next to your other channels

LinkedIn is the channel for B2B decision-makers. For consumers, Instagram usually works better; see the Instagram algorithm in 2026. If you would rather buy reach than build it, compare channels in how much do Facebook ads cost in Europe. The wider picture for next year is in social media trends 2027. And if you want your expertise to surface in AI assistants as well, read SEO vs GEO.

Images, video and AI in your posts

A post with your own photo or a short clip of your team holds a reader longer than a stock image. One shoot day gives material for months; see corporate photoshoot ideas. For video we shoot a corporate video and cut it into posts and reels. If you want a faster first draft, ChatGPT prompts for marketing help, as long as you supply the example and the number: exactly what a language model cannot know by itself. If you use AI-generated people or voices for an EU audience, check what applies from December in AI-generated content labelling. And if you prefer longer articles, the same rules apply as in how to write a blog post for SEO: a question, an answer and evidence from your own work.

How VBS writes your posts

From Ghent, Belgium, we run social media management for businesses across Europe: we pick three topics with you, write posts for your profile and your company page in Dutch, French or English, make the visuals and report on what they bring. If you want to be named when buyers ask an AI assistant about your field, we do that with found in ChatGPT. Want a LinkedIn plan for the Belgian market? Book a call.

A LinkedIn post every week that buyers read?

We choose your topics, write your posts and make the visuals, for your profile and your company page, in the languages of your market. Book a call.

Optional, if you prefer a call.

What do you want to achieve, and what is already there?

We only use your details to reply to you.