martin@tschapke.com LinkedIn Legal notice Privacy EN / DE
MRT Martin Reza Tschapke

Article

The LinkedIn algorithm now understands topics, and a clear profile makes its job easy

Since March 2026, a language model reads every LinkedIn post before it’s distributed. In five charts, I show how the new feed works, what LinkedIn itself says about it and how to set your topic so it lands.

Aug 27, 2026 · 4 min read
by Martin Reza Tschapke
A hand with a pen writing notes in an open notebook

If you post on LinkedIn, you’re hearing a lot about a new algorithm right now. Sometimes it’s called 360Brew, sometimes it’s said to punish posts written with AI. What’s really behind it, LinkedIn has described quite openly: on its engineering blog, in research papers and in posts by the man who leads the engineering behind the feed. The short version: LinkedIn now recognizes what a post is about. And it distributes it to the people who are interested in exactly that topic.

A language model reads every post

Since March 2026, LinkedIn has been rolling out a new system for the feed. Each post is described as text: format, post text, reactions so far and the author with headline, company and industry. A language model captures the meaning from this, not just individual keywords. LinkedIn gives its own example: someone interested in electrical engineering also sees posts about small modular reactors, even if the term electrical engineering never appears in them.

Process chain in four steps: the post is described as text, a language model captures the topic, pre-selection from hundreds of millions of posts, ranking for each individual member based on over 1,000 past interactions.
Simplified illustration based on LinkedIn’s engineering blog and two research papers.

A second model then ranks the selection for each individual member. It uses over 1,000 past interactions and, according to LinkedIn, only professional signals, no demographic ones.

Six sources become one

Until now, the feed drew on many pools: chronological posts from your network, trends by region and industry, recommendations based on similar members and a separate selection by topic. LinkedIn is now replacing these pools with a single pre-selection. For you, this means: whether a post is seen outside your network depends more on whether its content matches readers’ interests.

Before-and-after chart: until 2026, posts for the feed came from six sources, including chronological posts from the network, trends by region and selection by topic. Since March 2026, one pre-selection by a language model replaces these sources.
Source: LinkedIn Engineering Blog, March 2026.

And 360Brew? It’s a LinkedIn research model with 150 billion parameters, described in a paper published in early 2025. According to Tim Jurka, who leads the engineering behind LinkedIn’s feed, it was tested with a small group of members in early 2025 and then discontinued. It’s not part of the ranking.

Knowledge and advice have been favored for a while

The direction isn’t new. Back in 2023, LinkedIn shifted the feed towards content that offers knowledge or advice. According to LinkedIn, such posts then got almost 40 % more views from outside the network, and complaints about irrelevant content fell by 80 %. Editor in chief Dan Roth also explained what matters: LinkedIn wants to see someone building a community around knowledge they’re particularly qualified to share. If he posted about geology himself, LinkedIn would have to downgrade it, because he lacks the expertise.

Three key figures on the 2023 change to the LinkedIn feed: 40 percent more views from outside the network for posts with knowledge and advice, 10 percent more views from your own network, 80 percent fewer complaints about irrelevant content.
Figures from LinkedIn from 2023, before the switch to language models.

LinkedIn’s current help pages say something similar. The feed looks at three groups of signals: who you are, what your post is about and what your readers are interested in.

Three cards: your identity with profile, skills and professional background; your content, meaning what the post is about and whether it offers knowledge or advice; your readers’ activity, meaning what they’re interested in. Each with a recommendation.
Sources: LinkedIn Help and LinkedIn Engineering Blog.

What this means for your personal brand

There’s no figure proving how much reach a clear topic brings, not even from LinkedIn. But the logic is clear. If a language model is supposed to read from your profile and post what you stand for, and then distribute the post to interested people, every signal pointing in the same direction helps. If you write about sales today, parental leave tomorrow and AI tools the day after, you give the system little to hold on to.

This fits what LinkedIn announced in March 2026: less bait like “comment yes”, fewer off-topic videos and fewer recycled thought leadership posts. LinkedIn also wants to make groups that swap likes and comments ineffective.

LinkedIn analyst Richard van der Blom therefore recommends one main topic with two to three subtopics, together making up about 80 % of your posts. That’s a rule of thumb, not a measurement. As a planning aid, it still works well:

Example four-week content plan with three posts a week for a leadership coach: subtopics difficult conversations, the first leadership role and teams in transition, plus two personal posts.
Fictional example.

Where to start

  1. Your topic in one sentence. What should people who’ve never met you know you for? That sentence belongs in your headline.
  2. Align your profile. Write your About section, skills and experience so they tell the same story. Headline, company and industry are part of the description of every post you publish.
  3. Define two to three subtopics. Turn them into a four-week plan. Personal posts are welcome, as a complement.
  4. Review after four weeks. Which subtopics reach people outside your network? Keep going there and drop the rest.

I help self-employed professionals and coaches turn their knowledge into a clear personal brand, from the profile to the content plan. If you want to know what that could look like for you, get in touch.

Sources

  1. Danchev, H. (2026): Engineering the next generation of LinkedIn’s Feed. LinkedIn Engineering Blog, 12 March 2026. linkedin.com
  2. Jurka, T. (2026): Updates to The LinkedIn Feed Focusing on Authentic, Relevant Conversations. LinkedIn, 12 March 2026. linkedin.com
  3. Jurka, T. (2025): How does the LinkedIn Feed work? LinkedIn, 11 August 2025. linkedin.com
  4. Jurka, T. (2026): Post on 360Brew, LinkedIn, 14 April 2026. linkedin.com
  5. LinkedIn Help: LinkedIn relevance – Optimizing the member experience. linkedin.com
  6. LinkedIn (2025): Large Scale Retrieval for the LinkedIn Feed using Causal Language Models. arXiv 2510.14223. arxiv.org
  7. LinkedIn (2024): LinkedIn Post Embeddings: Industrial Scale Embedding Generation and Usage across LinkedIn. arXiv 2405.11344. arxiv.org
  8. Feifer, J. (2023): LinkedIn Changed Its Algorithms. Here’s How Your Posts Will Be Affected. Entrepreneur, 26 June 2023. entrepreneur.com
  9. Creator Science Podcast (2026): Richard van der Blom, The state of LinkedIn in 2026, 2 June 2026. creatorscience.com

More articles