All In
FR EN
Back to blog
Algorithme & stratégie LinkedIn 15 min read

LinkedIn H1 2026: What Six Months of Data Actually Show

| By Patrick de Carvalho

Contents

I run All In's editorial calendar in two languages, which means I read the same six months of LinkedIn data twice this spring: once assembled for a French audience, once for this piece. LinkedIn's first half of 2026 is the stretch from January through June during which the platform's 360Brew ranking model settled into place, the average engagement rate moved to 5.20%, and organic reach kept sliding by most published measures. This recap sources every number to a named study, states plainly where two sources disagree, and tells you upfront which figures are genuinely country-specific and which ones are global data wearing a local label.

In short: LinkedIn's average engagement rate reached 5.20%, up 8% year over year (Social Insider, 2026), even as several analyses put organic reach at 8 to 12% of followers, down from 15 to 20% a year earlier. Native documents, LinkedIn's term for carousels, led every format at 7.00% engagement, ahead of video at 6.00%. The 360Brew ranking model, a 150-billion-parameter system LinkedIn's own researchers described in a since-withdrawn arXiv paper, is the mechanism behind both numbers moving at once.

What changed on LinkedIn between January and June 2026

LinkedIn entered 2026 running on 360Brew, the ranking system its research team described in a paper posted to arXiv on January 27, 2025: a single 150-billion-parameter decoder-only model handling more than 30 ranking and recommendation tasks that used to require separate, hand-tuned systems. One detail rarely makes it into the marketing recaps, so I'll state it plainly: the paper was later withdrawn from arXiv, and the stated reason was a licensing issue on the submitter's side, not a retraction of the findings. I checked the withdrawal notice before citing the architecture here. The 150-billion-parameter figure and the "one model instead of many" design are the parts that matter for anyone posting, and they hold up.

Practically, that shift means a single system now reads a post's content, the author's declared profile, and the reaction the post gets, then decides distribution from all three inputs at once instead of running three separate scoring passes. What a profile says about you (job title, the "About" section, past roles) gets checked against what you actually publish. And the platform stopped needing to guess as much: it can score a post as engagement comes in, rather than locking in a verdict from a fixed early classifier.

Scale, for context on the base this all runs against: LinkedIn reports roughly 1.3 billion registered members worldwide. The United States remains the single largest national market at 239 million users, ahead of a combined Asia-Pacific base of 343 million and a European base of 314 million (Buffer, 2026). No public breakdown separates US-only engagement or reach figures from that global pool, a gap I return to later in this piece.

The format leaderboard six months of data actually support

Social Insider's 2026 LinkedIn benchmark report puts the platform-wide average engagement rate at 5.20%, up 8% year over year. Broken down by format: native documents lead at 7.00%, up 14%. Multi-image posts follow at 6.45%. Video sits at 6.00%, up 7%. Single images come in at 5.30%, up 9%. Text-only posts reach 4.50%, up 12% despite the smallest raw base. Polls land at 4.20%. Posts carrying an outbound link trail the field at 3.25%.

Format Engagement rate (2026) Year-over-year
Native document (carousel) 7.00% +14%
Multi-image 6.45% Not disclosed
Video 6.00% +7%
Single image 5.30% +9%
Text only 4.50% +12%
Poll 4.20% Not disclosed
Post with external link 3.25% Declining

Source: Social Insider, "LinkedIn Benchmarks 2026."

Buffer's 2026 analysis, working from a separate dataset, lands on the same order through a different method: carousels generate 278% more engagement than video and 303% more than single images, and video itself still beats a plain image post by 7% and a text post by 84%. Two firms, two methodologies, one ranking: document, video, image, text, link.

Metricool's 2026 study, built from 39 million posts across more than a million accounts, doesn't publish its format breakdown the way the other two do, but its headline finding, slower posting and denser formats winning the attention that's left, points the same direction.

Video's growth is real, and it's worth separating from the leaderboard because raw growth and relative engagement tell different stories. Buffer reports video output roughly doubling compared with other formats over the period, views up 36% year over year, building on an earlier six-fold quarter-over-quarter spike. LinkedIn Live, a smaller category, pulls seven times the reactions and 24 times the comments of a standard uploaded video. Video is growing faster than it's winning. That distinction got lost in most of what I read on this topic this spring, so I'm keeping it explicit here.

The link penalty deserves one line rather than a paragraph: a post carrying a URL in the body still trails every other format by a wide margin. Dropping the link into a first comment instead has been standard practice since well before 2026, and it still works.

Reach fell while engagement rate rose: the math behind the contradiction

Here is where the published numbers start arguing with each other, and I'd rather show you the disagreement than paper over it.

Social Insider's engagement rate climbed 8% year over year. A separate aggregation, published by UpGrowth in 2026 and drawing on research credited to Richard van der Blom and AuthoredUp, reports average post reach at 8 to 12% of followers, down from 15 to 20% a year earlier, with company page reach falling further, to 1.6% of followers from 7% in 2021. Read quickly, that looks like two sources disagreeing on whether LinkedIn got better or worse for publishers in H1 2026.

It didn't, and the reconciliation is the actual story. Reach is the size of the audience a post is shown to. Engagement rate is the share of that audience that reacts. 360Brew's stated design goal is to distribute fewer posts more precisely rather than spray every post to a fixed percentage of followers. Show a post to fewer, better-matched people, and the share who save, comment or click can rise even as the absolute number who ever see it falls. Both trends are consistent with one mechanism, not two.

I want to name the one figure in this pair I can't independently confirm. UpGrowth's write-up also states visibility fell 47% and engagement fell 39% year over year, attributed to the same van der Blom and AuthoredUp research without a citation I could trace to a published report. I'm including the reach-percentage range because two aggregators converge on it. I'm leaving the 47% and 39% figures out of the summary at the top of this piece and flagging them here instead, because I couldn't verify them against a primary source, and I'm not going to repeat a number I can't stand behind.

The practical read for anyone posting: saves, substantive comments (running 15 or more words, per the same reporting) and message shares now carry disproportionate weight against a like, on the order of five to ten times for a save and roughly fifteen times for a real comment, according to UpGrowth's synthesis, a hierarchy we break down at length elsewhere. I'd treat the exact multipliers as directional rather than precise. LinkedIn hasn't published its own weighting. But the direction, saves and genuine comments outweighing reflexive likes, is consistent across every source checked for this piece, including LinkedIn's own public description of what 360Brew evaluates.

Where B2B industries stand, sector by sector

Ghost.io's 2026 industry benchmark, built from 2.3 million posts across 47 sectors, gives the clearest sector-level picture available.

Sector Average engagement Top decile Year-over-year
SaaS & Tech 3.2% 8.1% Not disclosed
Professional Services 2.8% 7.3% Not disclosed
Finance & Fintech 2.6% 6.8% +0.7 pt (from 1.9%)
Marketing & Agencies 2.4% 9.2% Not disclosed
Manufacturing & Industrial 1.9% 5.7% Not disclosed

Median across all 47 sectors: 2.1%, up from 1.8% in 2025. Source: Ghost.io, 2026.

Two things stand out. Finance and fintech posted the largest confirmed year-over-year jump, which tracks with a genuine surge in posts explaining new compliance and reporting rules, a category readers search for rather than scroll past. And the gap between a sector's average and its top decile stays wide everywhere, widest in marketing and agencies (2.4% average against 9.2% for the top tier). In a sector whose entire pitch is expertise in exactly this, the distance between the median account and the best account is the least forgivable.

None of these five rows carry a country tag in Ghost.io's own reporting. I'm presenting them as global B2B benchmarks, not US-specific ones, because that's what the source actually says.

Newsletters, and the subscriber number I'm not going to repeat

LinkedIn newsletters keep winning on one metric that matters more than most: delivery. Moburst's 2026 review puts the average open rate for a company newsletter at roughly 40%, credited to LinkedIn's triple-notification system (email, push and in-app alert) against a standard feed post that reaches only 5 to 7% of followers on its first pass. More than 36,000 newsletters publish monthly on the platform, by the same count.

Here's the number I'm flagging rather than repeating: Moburst cites "over 500,000" active newsletter subscribers platform-wide. Far larger cumulative figures have circulated this year without a source I could trace to a primary report, and I'm not going to launder an unverified number by attaching a footnote to it. Five hundred thousand, sourced to one named study, is the figure I'll stand behind. If a bigger, better-sourced number surfaces, this piece gets updated, and I'll say so.

What the mechanism explains, regardless of which subscriber count you trust, is why the format works: a newsletter issue reaches its list through channels the ranking model doesn't touch at all. It sidesteps the entire reach conversation in the two sections above. For a B2B leader publishing on a topic with staying power, analysis rather than news, a monthly or twice-monthly newsletter functions closer to owned media than to a LinkedIn post, and it should be planned as such. We cover how to build that owned channel without leaning on the feed at all in our newsletter strategy guide.

What the French edition of this recap taught me about "country" benchmarks

Building the French-language version of this recap in parallel taught me something the marketing blogs don't say out loud: most of the "country" LinkedIn benchmarks in circulation are the same global studies with a local flag pasted on. Social Insider, Ghost.io, Buffer and Metricool all publish global or multi-country aggregates. None of the four breaks its engagement or format numbers out by country in the reports I could access. A "France 2026" or "US 2026" recap built on those four names is, more often than not, the same spreadsheet with a different byline.

What genuinely is country-specific, and worth naming instead of hiding behind a borrowed global average, is membership and behavior data published by local outlets tracking their own market. Blog du Modérateur, a French outlet, reports 37 million registered LinkedIn members in France and 13.5 million monthly active users, a 36% activity ratio, a roughly even 52/48 gender split, and a core user base between 25 and 34. That's a real, France-only data point. It tells a US reader almost nothing about a Chicago sales director's feed, and I wouldn't cite it here except to draw the contrast.

The practical takeaway if you're comparing benchmark posts: when a recap cites Social Insider, Buffer, Ghost.io or a similarly global firm and labels the finding "US" or "France," check whether the source itself made that distinction. Usually it didn't. The country label was added downstream, by whoever wrote the recap.

What I don't know

I don't have a single figure, from any source I could verify, that breaks LinkedIn's H1 2026 engagement or reach data out for the United States specifically. Every benchmark cited in this piece is global or multi-country. The 239 million US users figure is a membership count, not a performance number.

I don't know which of the two reach ranges is closer to reality, UpGrowth's 8 to 12% or Moburst's 5 to 7% for standard posts. Different measurement windows and different sample sources probably explain the gap, but I can't prove that from what's published.

I don't know if the 47% visibility drop and 39% engagement drop attributed to van der Blom's and AuthoredUp's research are accurate. I couldn't trace them to a report I could read directly, so I'm naming them as unverified rather than dropping them silently.

I don't know LinkedIn's own weighting of saves against likes and comments inside 360Brew. Every multiplier in this piece, five to ten times for a save, roughly fifteen times for a substantive comment, comes from third-party reconstruction, not from LinkedIn.

Five moves for the second half of 2026

None of this is new advice dressed up as a 2026 discovery. It's what the numbers above support, stated as five actions.

Lead with a native document over a text post when the choice is yours. The engagement gap, 7.00% against 4.50%, or 278% by Buffer's separate count, is too consistent across sources to argue with.

Build for the save, not the like. A resource, a comparison table, a checklist someone will reopen next month is what a save-weighted ranking system is built to reward.

Put the link in the first comment, every time. The penalty on an in-body link hasn't moved and shows no sign of moving.

Treat a newsletter as a second channel, not a repackaged feed. It reaches its list through notifications the ranking model never touches.

Ask what a benchmark actually measured before repeating its country label. Building this piece taught me that most of the "US-specific" and "France-specific" LinkedIn stats floating around this year are the same handful of global studies, relabeled. LinkedIn's own account of what changed with 360Brew, and how saves and sends fit into it, is covered in more depth on the All In blog. The engagement-manipulation crackdown LinkedIn escalated in March 2026, part of the same H1 story, is documented in our reporting on the platform's AI slop button, which covers a related enforcement push from later in the year.

FAQ

What was the average LinkedIn engagement rate in the first half of 2026?

5.20%, up 8% year over year, according to Social Insider's 2026 LinkedIn benchmark report. The figure is a global average across formats and industries, not a country-specific number.

Which LinkedIn format performed best in H1 2026?

Native documents, LinkedIn's term for carousels, led at 7.00% engagement (Social Insider, 2026), a result Buffer's separate 2026 analysis corroborates by a different method: carousels generated 278% more engagement than video and 303% more than single images.

Why did LinkedIn's reach fall while engagement rates rose?

Reach measures how many people see a post; engagement rate measures what share of them react. The 360Brew ranking model distributes posts to a smaller, better-matched audience rather than a fixed percentage of followers, which can lower absolute reach while raising the engagement rate among the people who do see the post.

What is 360Brew?

360Brew is LinkedIn's ranking system, described by its own research team as a 150-billion-parameter decoder-only model handling more than 30 ranking and recommendation tasks in one system instead of many separate ones. The description was published to arXiv in January 2025 and later withdrawn over a licensing issue, not a retraction of the findings.

Is a LinkedIn newsletter worth launching in 2026?

For topics with staying power, likely yes. Newsletters average around a 40% open rate (Moburst, 2026) through LinkedIn's email, push and in-app notifications, a delivery path that bypasses the feed-ranking system standard posts depend on.

Do LinkedIn benchmark reports actually break data out by country?

Rarely. The major firms cited in this piece, Social Insider, Buffer, Ghost.io and Metricool, publish global or multi-country aggregates without country-level engagement breakdowns. Country labels attached to their numbers in other recaps are usually added downstream, not sourced from the original study.

Does LinkedIn's algorithm reward saves more than likes?

Directionally, yes. LinkedIn has stated publicly that saves and substantive comments carry more weight than likes under 360Brew. Third-party reconstructions put the multiplier at roughly five to ten times for a save, though LinkedIn has not published its own exact weighting.

Sources

All In: the LinkedIn numbers, with the receipts attached

Every benchmark in this recap is sourced by name, and the two figures nobody could verify are said plainly instead of smoothed over. That's the standard the rest of the site runs on too.

All In is the B2B media that decodes LinkedIn, expert blog, weekly podcast and newsletter for SME leaders and sales directors who want to turn LinkedIn into measurable growth. An original creation by Patrick de Carvalho, on LinkedIn since 2004. Motto: "I Never Lose."

Discover All In and get what matters on LinkedIn, every week.