LinkedIn's 360Brew Rewrote the Feed in March. Reach Hasn't Bounced Back.

LinkedIn's 150-billion-parameter AI ranking system, 360Brew, has governed the main feed since March 12, 2026 — and creator reach has not recovered in the six months since.

Built on LLaMA 3 and fine-tuned on LinkedIn's data, it replaced a fragmented rule-based system. LinkedIn's Engineering Blog confirmed the March 12 deployment; the underlying research paper appeared in January 2025. The same architecture now handles feed ranking, jobs, ad targeting, and connection suggestions — over 30 predictive tasks at once, per LinkedIn's documentation.

What the rollout actually changed

The practical effect is not more reach. It is narrower, more relevant reach.

LinkedIn's own framing centers on "precision delivery": content routes to people most likely to find it useful rather than broadcasting to followers. Third-party data backs this up. Richard van der Blom's Algorithm Insights Report 2025 — which analyzed roughly 1.8 million posts across about 400,000 profiles — found views down 47 percent, engagement down 39 percent, and follower growth down 42 percent.

Three structural forces stack on top of each other: more creators competing for the same feed slots, paid ads claiming more real estate, and 360Brew's semantic matching shrinking each post's audience.

What does this change for marketers?

Benchmarks from 2024 do not apply anymore.

The depth-over-volume framework means saves, dwell time, comment depth, and private DM shares carry the algorithmic weight that likes used to. Per van der Blom's analysis, saves weigh 5× a like and 2× a comment. Posts that hold readers for 61+ seconds average 15.6 percent engagement; posts skimmed in under 3 seconds land at 1.2 percent.

Document and carousel posts lead at roughly 6.60 percent engagement, driven by the dwell time swiping generates. External links in post bodies carry an estimated 60 percent reach penalty. The "link in comments" workaround was patched in early 2026; LinkedIn now flags the bridge behavior directly.

What LinkedIn confirms vs. what creators claim

September 2026 brought no new confirmed ranking or distribution change. LinkedIn's own activity last month was product updates and marketing — not a feed-level mechanic shift. Anyone telling you "the algorithm changed again in September" is over-reading a test.

The active penalties are real. LinkedIn declared engagement pods a Terms of Service violation. LinkedIn banned Lempod, then the most widely used pod tool, and Google removed it from the Chrome Web Store. One practitioner-cited case saw reach collapse from 8,500 to 340 impressions after a pod flag.

AI-generated content with no original perspective takes an estimated 30 percent reach cut and 55 percent engagement cut, per third-party tracking. The trigger is not "you used an AI tool" — it is "you published content with no human professional perspective." Worth flagging because the line between the two stays blurry, and LinkedIn does not publish its detection thresholds.

What the source does not tell us

A few caveats matter before you rebuild your strategy on these numbers.

LinkedIn has not published its own reach or engagement figures since 360Brew deployed. Almost every statistic cited above comes from third-party analysis, primarily van der Blom's report. That dataset is large, but it remains one firm's analysis of observable behavior, not platform-confirmed distribution.

The 97 percent pod detection accuracy, the 30 percent AI penalty, and the 60 percent link penalty all come from third-party tracking. LinkedIn has not confirmed any of those multipliers. Treat them as directional, not gospel.

The "Depth Score" framing comes from practitioner interpretation of LinkedIn's signals, not from a LinkedIn announcement. Saves, dwell time, comment depth, and private shares are confirmed to be measured and weighted. LinkedIn added Saves and Sends to post analytics in late 2025 — about as close to an official steer as the platform offers.

The bottom line

360Brew is the feed now. Hooks, hashtags, and link-friendly formatting are not just fading; they are actively counterproductive. Post native insight, write for saves and substantive comments, keep external links out of the post body, and expect each post to reach a smaller but more qualified slice of your audience than it did two years ago.

Watch LinkedIn's Engineering Blog and press channels, not the algorithm-forecaster accounts. The next confirmed ranking shift will surface there first — and so far, six months in, nothing has.