Skip to main content
Growth

The LinkedIn algorithm in 2026: why your reach is falling, and what actually fixes it

LinkedIn reach is down across the board in 2026. Here is what 360Brew, dwell time, and the AI-slop crackdown changed, and the four levers that still grow reach.

By The Blendin team · Founders, Blendin14 min read
Blendin blog featured image: 'Why your reach is falling in 2026', a LinkedIn reach line that drops then recovers, over 360Brew, dwell time, and AI-slop labels.

Your posting has not gotten worse. Your numbers have. The same kind of post that pulled a few thousand views last year now stalls in the hundreds, the dashboard still says you published on schedule, and the quiet conclusion creeping in is that you did something wrong. You probably did not. LinkedIn rebuilt the machine that decides who sees what, and the new machine rations reach on rules most people on the platform have not read yet.

Here is the short answer, for anyone who only has thirty seconds. Reach is down because LinkedIn moved to a single large AI model that ranks the feed on relevance instead of raw popularity, and because the platform started actively limiting the reach of generic, low-effort content. You cannot reverse that. You can line up with it. This piece is the long version: what changed, why, and the four levers that still move reach in 2026.

Yes, your reach really is down

Start by naming the thing, because gaslighting yourself is a waste of a good Tuesday. The decline is real and it is broad. By one of the most widely cited measures, Richard van der Blom's annual Algorithm Insights report, average post views fell roughly 50% year over year, with engagement and follower growth down sharply alongside them. Almost every creator who is not in denial has watched the same curve bend on their own profile.

Now the part that reframes the whole problem. For most people, this was deliberate, not a punishment. The newer reporting on the same datasets shows that while raw views fell, engagement per post that does get distributed actually held up or rose, because the platform is now showing each post to a smaller, more relevant slice of the feed. LinkedIn would rather put your post in front of five hundred people who care about the topic than five thousand who scroll past. The reach you lost was mostly the reach that was never going to do anything for you.

That distinction matters because it tells you what to optimize. Chasing the old vanity reach is chasing a number the platform deliberately retired. The game now is relevance and retention: being unmistakably worth the slot you get. To understand how to do that, you have to understand the model making the call.

Meet 360Brew, the model behind the feed

For years the LinkedIn feed ran on a stack of separate, narrow ranking systems stitched together, each scoring a different signal. In early 2025 LinkedIn's research team published a paper describing 360Brew, a single large language model built to replace that patchwork with one unified ranker across the platform's surfaces. It is a foundation model, the same broad family of technology as the chat assistants you already use, pointed at the job of deciding what belongs in your feed.

The practical shift is the important bit. A model like this does not just count clicks and hashtags. It reads natural language, your profile, your post, the post's topic, the reader's history, and infers whether the two actually match. If a profile reads as a supply-chain consultant and the post is a generic motivational quote, the model sees the mismatch and holds the post back. If a RevOps leader posts a specific, useful take on pipeline forecasting, the model recognizes the alignment and pushes it to the RevOps people most likely to care. The feed went from popularity contest to relevance match.

Two things follow directly, and they will come up again and again below. First, coherence pays: a profile and a posting history that clearly stand for something give the model an easy, confident match to make. Second, specificity pays: a post about one real thing is legible to the model in a way that a post about everything and nothing is not. Pick a lane and write inside it.

Dwell time is the metric that matters now

If relevance decides whether your post enters a feed, dwell time decides whether it keeps going. Dwell time is simply how long someone's attention rests on your post: do they stop, expand the see more, read, swipe through the slides, or do they flick past in under three seconds. It has quietly become the cleanest quality signal the platform has, because it is hard to fake and it correlates with the thing LinkedIn actually wants, which is people finding the feed worth their time.

The gap is not subtle. By van der Blom's measurements, posts that hold attention for around a minute can see on the order of ten times the engagement of posts skimmed in a few seconds. The first line and the format are doing most of that work. A flat opener and a wall of text get the three-second flick. A sharp first line, a specific claim, and a format that invites a swipe buy the seconds that tell the algorithm to keep distributing.

This is also exactly where generic AI writing goes to die, which is why it is worth reading our companion piece on the tells that make writing sound like AI. A post that reads like a model wrote it gives a reader nothing to stop on, so dwell time collapses, so distribution stops. Nobody flags a file anywhere. The post just quietly fails to earn the one signal the feed runs on.

The golden hour, and why comments win

Timing still matters, just not in the way the best-time-to-post charts promised. What the evidence supports is a window, often called the golden hour, of roughly the first thirty to sixty minutes after you publish. Strong early engagement in that window signals the algorithm to widen distribution; a flat start tells it to stop. This is why the same post can travel one day and sink the next depending on nothing more than whether anyone engaged early.

Inside that window, not all engagement is equal. Comments carry far more weight than likes, and substantive comments, the kind with a real sentence in them, carry more than a one-word “great post.” The reason ties back to dwell time and relevance: a thoughtful comment is proof a real person stopped and thought. So after you publish, the best move is not another post. Spend that first hour in the comments, on your own post and on other people's, having real conversations.

One honest, practical note on this, because it is where a lot of tools overpromise. LinkedIn down-ranks links in the body of a post, so the common move is to put your link in the first comment. Blendin will draft that first comment for you, ready to paste the second your post goes live so it lands inside the golden hour. You paste it. It is never auto-posted on your behalf, because a human in the loop is the entire point of staying on the right side of the platform.

The AI-slop crackdown

In 2026 LinkedIn went from implicit to explicit. The platform announced measures to limit the reach of low-quality, AI-generated content, the stuff the industry now calls AI slop: posts that are fluent, generic, and empty of any original perspective. Coverage of the change framed it plainly, LinkedIn wants less of the machine-written filler clogging the feed, and it is willing to suppress reach to get there.

Here is the nuance that most hot takes miss. The algorithm is not a reliable AI detector, and LinkedIn has not claimed it is one. It does not need to be. It reads behavior. Generic AI content earns near-zero dwell time, no saves, and no real comments, and the model reads that behavioral fingerprint and quietly stops distributing the post. Whether a human or a machine typed it is almost beside the point. The post failed because no one cared, and AI slop is engineered, by accident, to be the thing no one cares about.

So the takeaway is not “hide that you used AI.” It is “do not publish anything, by any method, that reads like the average.” If you want a fast gut check before you post, run the draft through our free AI Writing Checker, which flags the same tells the feed's readers spot in three seconds.

The four levers that still grow reach

Strip away the model names and the panic, and what actually moves reach in 2026 is short, boring, and durable. Four levers, in rough order of how much they matter.

1. Consistency. The single most reliable signal is that you show up. Accounts that publish two to five times a week earn an order-of-magnitude more impressions per post than accounts that publish once a week or less, and the model treats a reliable cadence as evidence you are a creator worth distributing. We made the full case, with the math, in the essay on why only 2% of LinkedIn publishes weekly. Consistency is the lever that makes the other three compound.

2. Sound like a person. Dwell time is the new currency, and nothing kills dwell faster than prose that reads generated. In 2026, sounding human is a reach lever, not just a style preference.

3. Pick a lane. 360Brew rewards a clear topic-to-profile match. A feed of posts that all orbit one or two themes is far easier for the model to place than a scattershot of unrelated takes. Narrow beats broad.

4. Use formats that earn dwell. Some formats are simply better at buying the seconds that drive distribution. That is the next section.

The formats that earn dwell time

If dwell time is the metric, the formats that win are the ones that ask for more than a glance. Carousels and native LinkedIn Documents are the clearest example: every swipe is another second of attention, and the platform reads that as quality. By van der Blom's data, document posts have been among the highest-engagement formats on LinkedIn, outperforming plain text by a wide margin, precisely because a reader who swipes through eight pages has spent real time with your idea.

The catch is the one that keeps most people on text-only posting: building a carousel or a multi-page Document by hand is slow. A good one is hours of layout, type, and color work per post, which is not a budget anyone sustains at a two-to-five-times-a-week cadence. That is the bottleneck worth removing, because it is the difference between knowing carousels work and actually publishing them every week.

What to do in the next 30 days

You cannot reverse-engineer 360Brew, and you should not try. You can line up with how it works. A month-long plan that does exactly that:

  • Pick one lane. Choose the single topic you want the model to associate with your profile, and make sure your headline and About section say it plainly. Coherence is a ranking input now.
  • Commit to a scheduled cadence. Two to three posts a week, produced in one sitting and scheduled, not improvised each morning. Reliability beats volume.
  • Lead with a specific first line.The opener buys the dwell time that buys the distribution. Name the number, the client, the thing that went wrong. Skip the “I am thrilled to announce.”
  • Protect the golden hour. Publish when you can be present for the next hour, then go reply to comments like a person, not a brand.
  • Run every draft through a tells check. Generic prose is now a reach tax. The free AI Writing Checker takes ten seconds.
  • Favor carousels and Documents. Make at least one swipe-based post a week. It is your best dwell-time bet.
  • Read the post-level results, then double down. Blendin's Insights show how the posts you publish with Blendin performed, post by post, so you repeat what worked instead of guessing. It is post-level performance, not audience dashboards you will never open.

None of it is a hack. The new algorithm just rewards unglamorous, compounding behavior, run on a schedule you can actually keep. For the deeper version of the cadence half of this plan, the 2% essay has the full 30-day build.

The bottom line

The 2026 reach drop is real, but it is not a glitch and it is not aimed at you personally. LinkedIn replaced a popularity engine with a relevance engine, 360Brew, and started actively starving the generic content that was clogging the feed. The platform stopped paying for volume and started paying for attention. Reach did not disappear. It got rationed, and the currency is whether a real person stops and reads.

You cannot game that. You can earn it, with four unglamorous moves: show up on a schedule, sound like a human, pick a lane, and use the formats that hold attention. Do those four for ninety days and the curve that bent against you starts bending back. That is the whole strategy, and it happens to be exactly what Blendin was built to make easy.

Sources and further reading

Third-party figures in this piece are attributed inline and linked below. Verify the latest against the original sources before citing them elsewhere; the platform and the research both move fast.

Want the rest of the cluster? Read why only 2% of LinkedIn publishes weekly for the cadence case, and the AI writing tells that quietly cost you reach for the dwell-time half. Or compare Blendin plans and the free monthly credits and start publishing on a schedule today.