Social Rocket
LinkedIn distributes posts more deliberately to people who do not follow their authors. On November 9, 2026, the network described in detail for the first time how its new Profile Relevance model matches professional profiles and posts using language models. In a live test, the rate of hidden posts dropped by 18,8 percent. For B2B brands and expert creators, the real news lies in the selection principle: professional relevance becomes an additional path out of your own contact bubble.
LinkedIn reads not just likes, but professional context
LinkedIn's Engineering team describes a search and recommendation component for “Unconnected Content” – posts from people outside your own network. To do this, the model converts the member's job title, skills, positions and education into a semantic representation. LinkedIn sets this against the content of posts: text, title, transcripts where applicable, and information about the author.
This is not about a simple list of matching keywords. A post about industrial robotics, for example, may be interesting to people in a relevant technical job, even if they don't know the author yet.
The new infrastructure complements a broader LLM-based feed ranking that LinkedIn had already introduced in March. Back then, it also concerned the timing of interactions. Profile Relevance now explicitly addresses the discovery of content beyond existing connections.
A first signal from outside the network counts
What's interesting for social teams is a detail in LinkedIn's technical report: profile matching alone is not enough. For distribution via Profile Relevance, LinkedIn requires at least one interaction from an unconnected person – such as a like, a comment or a share. The system thus combines topical fit with a small piece of evidence that a post also resonates beyond existing contacts.
Geography matters too. In tests, regional signals were partly lost in a heavily shortened text representation. LinkedIn then added country information from members and post creators so that local professional topics find each other more reliably. For German-language expert communication in particular, this is noteworthy: a well-matched local context can play a role alongside industry and job profile.
What the tests actually show
According to LinkedIn, several variants were compared online and offline. The variant used in production brought the following changes compared with the previous system:
- 18,8 percent fewer hidden posts
- 0,54 percent more viewed feed posts
- 0,94 percent higher “Viral Rate”
- For members with little usage history, 17,7 percent more viral engagement and 1,8 percent more daily use
LinkedIn's platform tests make the optimization visible: users should discover topically relevant posts beyond their network without receiving more irrelevant content in their feed.
How B2B marketing should plan differently now
For corporate influencers, personal brands and company pages, three practical approaches emerge. First, posts should address their topic concretely enough that readers can immediately see the professional benefit. Second, it is worth checking whether author profiles accurately describe expertise and role – the profile is a data signal in the recommendation model. Third, teams should not only track follower counts but also reach, profile visits and qualified reactions outside their existing network.
The mechanism fits LinkedIn's growing creator ambitions: according to Srocket's latest analysis, LinkedIn creator ads rose by 63 percent. At the same time, advertisers can use Creator Discovery to find suitable people for B2B campaigns. Anyone building an organic professional audience can thereby potentially also create an environment for credible collaborations; the Creator Marketplace is, however, still limited in Germany.
The strategic direction is clear: LinkedIn is trying to make professional relevance measurable beyond network size. This could give highly specialised voices more discovery.
Image: LinkedIn Sales Solutions / Unsplash.