LinkedIn now ranks post comments by their relevance to each individual viewer rather than by when they were posted, a change surfaced on Threads by creator-economy analyst Lindsey Gamble and reported August 9 by Andrew Hutchinson at Social Media Today. In LinkedIn’s own words, the system uses signals “including” “professional interests, connections, and engagement activity.” That single word, including, is doing most of the work: the platform declined to name the full signal set or disclose weightings.

The context is the point. LinkedIn’s Q2 2026 performance report showed an 18% year-over-year jump in time spent inside post comments and a 10% rise in overall content consumption. Then Pangram Labs, an AI-detection startup, published an analysis of 57,000 public posts and concluded that 30% of comments on the platform between April and June were entirely AI-generated. Engagement was up. Authorship was, increasingly, not human.

The relevance re-rank is the enforcement mechanism. Coupled with what Social Media Today describes as an active crackdown on engagement pods and AI spam, it dismantles the entire architecture of pod-based B2B growth: the coordinated early comments, the reciprocal likes, the race to sit at the top of a founder’s post before the algorithm calcifies. That ordering doesn’t exist anymore. Each reader sees a different comment stack.

Bernie Fussenegger, writing in B2The7’s September 14 marketing trends roundup, put the shift plainly: “the race to comment first matters less, the quality of what you say matters more.”

For small B2B teams, the practical implications are sharper than the platitude suggests. Vanity metrics like “our comment ranked second” are now analytically dead, because there’s no shared second. The signals worth watching are downstream: replies to your comment, profile visits, inbound connection requests. Volume plays lose. Substantive replies from operators with actual domain knowledge get algorithmically routed to the exact prospects who care.

Several things remain undisclosed, and they matter. LinkedIn hasn’t published a rollout timeline, named markets or languages, confirmed whether a chronological view survives, or clarified whether company pages behave differently. Pangram’s detection methodology, error rate, and sampling approach are also unpublished. Read alongside LinkedIn’s earlier AI-slop crackdown, the direction of travel is unmistakable, even if the machinery isn’t.

Sources