The short answer
Part of the answer is now public. On 9 November 2026, LinkedIn's engineering team explained Profile Relevance, a system that uses a large language model (LLM) to show members posts from outside their network. It reads each member's profile (headline, About section, skills, work history and education) and each post's text, title and video or audio transcript, then matches the two by topic, with the country added so location isn't lost. A post needs at least one like, comment or share from a member who isn't a connection before it's recommended. For a small business, a founder-led business or a freelancer, that means the words in your posts decide who outside your network they can be matched with.
What did LinkedIn publish?
Most of what gets written about the LinkedIn algorithm is guesswork from people watching their own numbers. This one comes from the engineers who built part of it, in an article on the LinkedIn Engineering blog called "Profile Relevance: boosting feed relevance with LLMs at LinkedIn".
The article explains that recommendation systems have relied heavily on who you're connected to and on what gets engagement, and that good posts also come from people outside your network. LinkedIn calls those posts Unconnected Content. Profile Relevance is a new way of finding them, and LinkedIn describes its launch as a key source of Unconnected Content in the feed.
What does it read about you?
It reads a summary of the professional information you chose to put on your public profile:
- your headline, your About section and your key skills
- your work history: companies, job titles, how long you were there and the role descriptions
- your education: schools, subjects and degrees
Your name and gender are left out. Your country, taken from the location on your profile, is added.
The team tested two versions. In one, a model reads your profile and pulls out your Top 10 Topics, and those topics are what gets matched. In the other, a model reads the profile text as it is, which keeps more detail. Either way, your profile becomes a description of what you work on, and the posts LinkedIn finds for you from outside your network are the ones closest to it.
What does it read about your post?
The post's text, its title, and the transcript of any video or audio. Details about the author, such as their country where it's available, are added too.
The article doesn't mention images. I wouldn't rely on words that only appear inside a graphic, so if the point of your post is on the image, say it in the text as well.
Location was one of the team's main findings. Summarising profiles into topics weakened the location signal, so they added the member's country and the author's country back in by name, and local relevance improved.
When does a post get recommended?
Only after it has at least one like, comment or share from a member who isn't a connection. Engagement works as a gate here: it decides whether a post can be recommended at all, and the match between profile and post decides who to. The match itself is worked out from the words on both sides, without click data.
Did it make a difference?
LinkedIn compared Profile Relevance with the source it had used before. In online tests, the version they settled on, which adds location and the engagement gate, showed:
- a 0.54% rise in feed updates viewed
- a 0.94% rise in viral rate
- an 18.8% fall in hide rate
For members with little activity history, the ones LinkedIn knows least about, viral engagement rose by 17.7% and daily active users by 1.8%. When someone hasn't liked or commented much yet, their profile is what LinkedIn has to go on.
Four things to do with this
1. Say what each post is about, in plain words
The text is what gets matched, so make the topic clear from the words alone. A post that names the subject ("how to get a café found on Google Maps") gives the system something to match. A post that says "Big news!" over a photo gives it very little.
2. Put your key words into what you say on video
Transcripts get read, so the words you say in a video count as much as the caption. Name the topic out loud near the start.
3. Keep your location on your profile
LinkedIn adds the author's country to each post and the member's country to each profile, because location got lost without it. If your location is missing, that part of the match is missing too.
4. Write your profile around the topics you work in
Your headline, About section and skills decide which posts from outside your network you're shown. If they describe your work clearly, you'll see more posts from people in your field you're not connected to yet, and those are good posts to comment on when you want to meet new people. One test: could a stranger read your headline and About section and name your top topics?
If you'd like help with your LinkedIn profile and posts alongside your Google profile and website, that's weeks two and three of Found & Chosen.
What the article doesn't say
This article will get quoted loosely, so I'd rather be clear about its edges. It doesn't say how much each part of your profile counts, so if someone tells you it proves your headline counts more than your About section, it doesn't. It doesn't name which of the two versions went live, or whether every member has it. And it covers one source of posts from outside your network, so it tells us nothing about the rest of the feed or the posts you see from your connections.
Does this help you in AI search too?
This part is my view, not something the article covers. ChatGPT, Perplexity and Google's AI answers also put together what they say about you from the words you've published, and a LinkedIn profile is one of the places they can read. So the same habits help in both places: say what you do, who for and where in plain words, and say it the same way everywhere. My guide on how to get your business to show up on ChatGPT covers the rest.
More questions about getting found? The getting-found FAQ on the homepage answers the ones I hear most, and the other guides are on the Guides page.
Jitka Lopez
Getting businesses found has been part of my job for 15+ years, from multi-market website migrations for Unilever's beauty brands across Asia to international SEO on Baidu, Yandex and Naver. LinkedIn and AI search are the newest versions of the same job: say clearly what you do, be consistent, and be talked about.
If you'd like to set all of this up properly with someone checking your work, Found & Chosen covers it in four weeks. If you'd rather learn it yourself alongside other founders, join the Northern AI Collective.
Sources
- LinkedIn Engineering, Profile Relevance: boosting feed relevance with LLMs at LinkedIn, by Shiyong Lin and colleagues, published 9 November 2026. Every fact and figure in this guide comes from this article.
