LinkedIn Algorithm 2026: How Posts Get Ranked and What to Post

The LinkedIn algorithm in 2026 is a large language model ranker called 360Brew that reads what your post means, matches it to each member's interests and predicts whether they'll stay on it ("long dwell") and whether they'll comment, like or share. It rewards people who post on the same two or three topics and start real conversations. It demotes engagement pods, automated comments, AI slop, engagement bait and thin link posts.

The short version

• LinkedIn's feed now runs on an LLM ranker that reads meaning and predicts dwell and conversation. Keyword and hashtag tricks don't move it.
• Stick to 2-3 core topics. Richard van der Blom's 2026 report links that focus over 90 days to +27% reach.
• Comments and saves beat reactions. Reply to every comment on your posts.
• Under 5K followers, images do better than carousels. Above 20K, documents take the reach lead.
• Aim for 1,300-2,500 characters, short paragraphs and a hook inside the first 140 characters.
• Expect less reach than in 2023. Van der Blom found reach down nearly 50% for 95% of creators.

How does the LinkedIn algorithm work in 2026?

LinkedIn replaced its old feature-scoring system with one large model. 360Brew is a 150-billion-parameter model, described in a research paper in January 2025 and reported live in the feed from March 12, 2026. It handles 30+ prediction tasks and feeds two to three months of each member's activity into its prompt before ranking anything (AuthoredUp).

Buffer interviewed LinkedIn staff for its 2026 guide. Their summary is that the feed weighs connections, topics and "the conversations they're most likely to join," and that virality is deprioritized: "it's more about meaningful connections and relevant conversations now." The same piece cites a LinkedIn engineering paper from early 2026 that describes predicting whether someone will stay on your post past a certain point, which is the long dwell signal.

Tim Jurka from LinkedIn has said the company now uses LLMs and larger sequence models to understand what a post is about and how a member's interests change over time. Hootsuite reads it the same way. The ranker tries to understand the meaning of a post instead of scanning for keywords, hashtags or exact phrases.

SocialPilot ties the March 2026 rebuild, which swapped the old multi-source setup for LLM retrieval and a transformer ranker, to the out-of-network reach drops many creators saw. Its wording is that the drops are "likely the result of this rebuild."

What signals does the LinkedIn algorithm reward?

LinkedIn doesn't publish weights. Everything numeric below comes from third-party analyses, and most of it is vendor data that's observational and correlational. Treat the multipliers as directional.

Dwell is the one you can't see. AuthoredUp calls it an internal signal creators have no view of, and says carousels score well partly because each swipe adds time.

The reply finding is the best-evidenced tactic here. Buffer used a fixed-effects model, and the effect was positive for about 83% of profiles. It's still correlational, since popular posts may attract more replies on their own. But no source found a downside, so reply to everything.

What is topic fingerprinting on LinkedIn?

Topic fingerprinting is van der Blom's name for how the ranker learns what you're an expert in. His 2026 thesis is that LinkedIn now distributes by interest more than by connections, a shift "from rewarding who you know to rewarding what you want to be known for."

His data puts a number on it. Sticking to 2-3 core topics for 90 days correlated with +27% reach and +41% topic-based follows (Kara Redman's summary).

LinkedIn's own creator guidance points the same way: members engage more with deep dives on niche topics than broad ones.

My take: pick your two or three topics and write them into your headline and About section. Profile hashtags are gone, and LinkedIn now infers expertise from keywords across About, Experience and Skills (SocialKit). Your profile and your posts should describe the same person. If you're posting around a day job, the full-time job guide covers how to pick topics that won't get you in trouble.

What gets your LinkedIn reach demoted?

LinkedIn has been explicit about pods and automation. The rest comes from vendor data.

Engagement pods. LinkedIn's newsroom says it has removed groups showing pod behavior and warned members using pods or automation tools, with possible loss of Top Voices status. A LinkedIn spokesperson said a post that's been part of a pod may be recommended less outside the author's network.

Automated comments. VP of Product Gyanda Sachdeva said comments posted via scripts or browser plugins are removed from the default "Most Relevant" view, and their reach is limited to the commenter's direct connections. Enforcement was described as arriving February to March 2026.

AI slop. SocialPilot reports that content LinkedIn classifies as slop gets about 40% fewer views, and another SocialPilot page says 30%. SocialPilot adds that using AI on its own doesn't trigger the demotion.

Engagement bait and hard sells. Posts that demand comments or DMs for a resource get suppressed. Postiz goes further: content full of external links, hard-selling language or CTA-heavy messaging is now treated like engagement bait. Taplio adds that "comment what you do" networking posts don't work either.

Plain reposts. A repost with no commentary gets little visibility. Add your own take on top.

Anything that fakes conversation or cuts dwell short costs you. That's also why hand-written comments matter more now. Van der Blom says about 80% of early comments on his posts are AI-generated, so a specific human comment stands out. The LinkedIn commenting guide covers what a good one looks like.

Do external links hurt LinkedIn reach?

The sources disagree, so here's the evidence and then a rule.

• Metricool's 577,000-post study found posts with links got 13.57% more interactions and 4.90% more views.
• Hootsuite's 2021 experiment found posts without links got 6x more reach.
• Ligo's 134,522-post dataset puts link previews at -15% engagement against the same author's text posts.
• ContentIn notes LinkedIn has never published a policy demoting posts for containing a link.

LinkedIn's own position, from Laura Lorenzetti via Buffer, is that for link posts the algorithm focuses on the value of the text written alongside the link.

My rule: a link is fine inside a post that stands on its own. A two-line "here's my new article" post is the link dump that gets punished. If you're unsure, Taplio and Postiz default to the first comment, and one 2026 writer suggests the second-to-last line of the post instead.

Which LinkedIn post formats get the most reach?

It depends on whether you want reach or engagement, and on how many followers you have.

Sources: Buffer State of Engagement 2026, AuthoredUp formats, AuthoredUp video, Creator Science.

Two warnings on the carousel number. A B2B case study argues carousel engagement is inflated because each slide click counts as an interaction. And AuthoredUp found carousel reach down 18% and engagement down 25% year over year.

Video splits the sources too. Buffer calls it "fantastic for reach and discoverability", and in Buffer's own 30-day test seven of the eight most-engaged posts were videos. Upload natively instead of pasting a YouTube link (Postiz).

My take by account size:

• Under 5K followers: image plus strong text as your default. Carousels for your best frameworks.
• 5K-20K: mix images and carousels.
• Over 20K: lean into documents and longer native video.

Whatever the size, judge each format against your own median.

How long should a LinkedIn post be in 2026?

AuthoredUp's sample of 372,126 personal-profile posts found the highest median engagement in the 1,301-2,000 character band (2.61%) and the 2,001-2,500 band (2.67%). That's roughly 200-400 words. The hard cap is 3,000 characters.

Formatting matters because dwell matters. A post split into 20+ short paragraphs lands at 1.13x reach and 1.23x engagement, which AuthoredUp calls the biggest single formatting lever.

The fold is about 140 characters on mobile and 210 on desktop (LinkedinPreview). So:

1. Line one carries a specific number or tension, under 140 characters.
2. Line two "rehooks" to earn the See more click.
3. Write 10-20 one- or two-line paragraphs with the lesson or story.
4. End with one specific question.
5. Write the hook last.

On content, first-hand beats generic. In Ligo's dataset, stories earned +38% engagement and first-hand-experience posts +41% against the same author's advice posts. For cadence, see how often to post on LinkedIn; van der Blom's 2026 number is 2-4 posts a week, down from 5-6.

Does editing or scheduling a LinkedIn post hurt reach?

No source documents an edit penalty. Magicpost says LinkedIn "does not document any reach penalty for editing a published post, and we have not measured one in our own data." LinkedinPreview says minor edits have zero impact.

One outlier, ConnectSafely, claims major edits after 30 minutes cut impressions 30-50%, but its sample is 340 posts with thin methodology (ConnectSafely). Fix typos freely. Avoid rewriting the whole post an hour in.

Scheduled posts aren't penalized.

Does the first hour still decide a LinkedIn post's reach?

Partly. Expandi, citing van der Blom, says 70% of total reach is decided in the first 60-90 minutes. Hootsuite says LinkedIn determines value within the first hour.

But the newer ranker also gives relevant posts a long tail. Hootsuite says LinkedIn confirmed it's more likely to show older posts, 2-3 weeks old, if they're relevant. Later says suggested posts resurface for months.

The rule that satisfies both camps: be online for the hour after you post and reply to every comment. Then stop worrying about the exact minute. Best-time data disagrees by up to ten hours between Sprout, Buffer and Hootsuite anyway, so pick a consistent weekday slot and check your own analytics.

How much reach should you expect on LinkedIn in 2026?

Less than you remember. Van der Blom's 2025 report covered 1.8M posts over the 12 months to February 2025 and found reach down nearly 50% for 95% of creators, engagement at 75% of prior levels and follower growth at 59%. His 2026 report puts reach down about 60% for active creators over two years.

If your impressions have halved since 2023, you're in line with the data. Measure yourself against your own median from the last 90 days.

The sources agree on what still works. Pick two or three topics, post 2-4 times a week, reply to every comment and spend 15 minutes a day writing real comments on other people's posts. The four-platform playbook shows how that fits next to X, Instagram and Substack.

FAQ

What is 360Brew on LinkedIn?

360Brew is LinkedIn's LLM-based ranking model, described in a January 2025 research paper and reported live in the feed from March 2026. It has about 150 billion parameters and reads two to three months of each member's activity to predict what they'll dwell on, comment on or share.

Do hashtags still work on LinkedIn in 2026?

Barely. LinkedIn dropped hashtag following and profile hashtags in 2024, and hashtags now act as a minor topic signal. Most guides suggest 1-5 relevant tags, but don't expect them to move reach.

Is the LinkedIn link penalty real?

The data conflicts. Metricool found links lifted interactions 13.57%, while Hootsuite and Ligo found link posts underperform. LinkedIn says what matters is the value of the text around the link, so put links in posts that stand on their own.

Why did my LinkedIn reach drop?

Most creators' reach fell. Van der Blom found reach down nearly 50% for 95% of creators in 2025, and SocialPilot links further drops to the March 2026 feed rebuild. Narrow your topics, reply to every comment and compare against your own recent median.