LinkedInLinkedIn

Why Your Reach on LinkedIn Decreased in 2026 (and What the Algorithm is Looking at Now)

Team at AiToolsForAppsWritten with AI assistance16 min read

In summary

  • The reach has really dropped, it's not just your impression. The 2026 LinkedIn study by Metricool (673,658 posts from 63,108 accounts) measures a decline in average impressions per post from 924.42 to 831.96, a -10% year-on-year.
  • On November 9, 2026 LinkedIn explained its Profile Relevance model: it uses embeddings generated by an LLM to measure how well each post fits with the professional context of the reader (topic, location, position, company, sector, seniority, and education).
  • The stated goal is relevance, not volume. In LinkedIn's experiments, the model reduced the times someone hides content by -18.8% and increased viral interaction among new members with little history by +17.7%.
  • Watch out for the 360Brew legend: it is not your feed's algorithm. LinkedIn's VP of engineering Tim Jurka confirmed on April 14, 2026 that it was an internal test and that they shut it down.
  • What does translate into work: document posts (the PDF carousel) are the format that retains best — 1.39× reach and 1.30× interaction over the average, according to AuthoredUp — and your headline and your "About" have become raw material for the ranking.

If you've been seeing the same posts that used to get 4,000 impressions now only getting 800, you're not paranoid and you haven't been "punished": LinkedIn has changed the question it asks before showing a post. This article gathers what the platform has officially stated in 2026, what independent studies measure, and what can be deduced from it for those using LinkedIn to sell services. In the end, you'll see why this changes day-to-day work, and where aiLinkedin fits, the box of AI tools for covers, headlines, carousels, and LinkedIn posts from AiToolsForApps.

Laptop screen with a flow of posts where almost all appear dimmed and only two stand out, over a mesh of connected nodes
Laptop screen with a flow of posts where almost all appear dimmed and only two stand out, over a mesh of connected nodes

Has LinkedIn's reach dropped in 2026 or is it just me?

It has dropped, and it is measured. The 2026 LinkedIn study by Metricool, based on 673,658 posts from 63,108 accounts, places average impressions per post at 831.96, down from 924.42 the previous year: a -10%. In its 2026 general social media study, published on December 10, 2025, covering nearly 40 million posts, the decline attributed to LinkedIn is even greater: -23% impressions and -14% interactions.

The interesting data is that interaction did not plummet at the same time. In that same LinkedIn study, the engagement rate rose from 12.21% to 13.90% and average clicks per post increased from 97.54 to 102.32 (+4.90%). In other words: fewer people see each post, but those who do react more. That is not the profile of a declining platform; it is the profile of a platform that shows each thing to fewer and better-chosen people.

Metric (LinkedIn)Before2026VariationSource
Average impressions per post924.42831.96-10%Metricool, LinkedIn study 2026
Engagement rate12.21%13.90%+1.69 p.p.Metricool, LinkedIn study 2026
Average clicks per post97.54102.32+4.90%Metricool, LinkedIn study 2026
Impressions (cross-platform aggregate)——-23%Metricool, social study 2026 (Dec 10, 2025)
Interactions (cross-platform aggregate)——-14%Metricool, social study 2026 (Dec 10, 2025)
Video reach, year-on-year——-36%AuthoredUp, Mar 2025 to Feb 2026

To this, we add the simplest and least commented factor: there are more people publishing. The CEO of Metricool, Juan Pablo Tejela, summarizes it in the study's statement saying that "algorithmic overcrowding is real." LinkedIn, for its part, stated to Entrepreneur on May 19, 2026, a 14% increase in content creation year-on-year. Same showcase, more product: each person has less display.

What LinkedIn Announced on November 9, 2026: the Profile Relevance Model

On November 9, 2026, LinkedIn published details of its Profile Relevance model on its engineering blog, presented by engineer Shiyong Lin and covered by Social Media Today. The model uses embeddings generated by an LLM to match members and posts based on professional context: in practice, LinkedIn builds a semantic summary of who you are professionally and uses it to decide what appears in your feed.

The signals that, according to the announcement, contribute to this relevance assessment are specific and all come from your profile or that of the reader:

  • Topic of the post
  • Geographic location
  • Job title
  • Company
  • Sector
  • Seniority
  • Education
A post connected with a strong thread to a matching professional profile and with weak threads to two profiles that don't fit
A post connected with a strong thread to a matching professional profile and with weak threads to two profiles that don't fit

The results that LinkedIn reports from its online experiments are small in percentage but very revealing in direction:

Experiment ResultVariation
Feed updates viewed+0.54 %
Viral rate+0.94 %
Times a user hides content-18.8 %
Viral interaction in new members with little history+17.7 %
Daily active users (DAU)+1.8 %

Notice what the big number is: -18.8 % of hides. The metric that LinkedIn has genuinely moved is not "how much engagement there is," it's "how much what we show annoys users." And the +17.7 % in new members with no history says the other important thing: when the system has no past behavior to draw upon, it reads the profile. That is the hinge of this entire article.

No, 360Brew is not your feed algorithm

Much of what you will read this year about "the new 360Brew algorithm with 150 billion parameters" is false, and it’s worth stating that before proceeding. 360Brew was a research model, described in a paper by LinkedIn authors published in 2025 and subsequently retracted from arXiv by its lead author. On April 14, 2026, Tim Jurka, VP of engineering at LinkedIn, explicitly stated that it is not used in the ranking system: they tested it with a small group of members, decided it did not fit the platform, and closed the trial. The journey of this rumor is documented in this denial about 360Brew.

This matters for a practical reason: half of the advice circulating is derived from a model that is not in production. And in the same family are the exact multipliers that you will see repeated in tool guides — "a save counts as 5 times a like," "pods are detected with 97 % accuracy," "generic AI loses 30 % of reach" —. These figures appear in commercial resources such as the SocialPilot LinkedIn algorithm guide, without LinkedIn ever having published them. The direction they indicate is reasonable; the specific number cannot be cited as a fact, and you certainly should not build your editorial calendar around it.

What you can take as true, because LinkedIn has said it, is this:

StatementConfirmed by LinkedIn?
There is a Profile Relevance model with LLM embeddings that matches posts and professional contextYes — engineering blog, Nov 9, 2026
The ranking considers job title, company, sector, seniority, education, topic, and locationYes — engineering blog, Nov 9, 2026
LinkedIn reduces the distribution of generic AI content and automated commentsYes — statements to Entrepreneur, May 19, 2026
360Brew is the feed algorithmNo — denial by Tim Jurka, Apr 14, 2026
A save counts exactly as 5 times a likeNot stated — commercial origin figure

Why generic AI text performs worse and worse

Because LinkedIn has openly stated that it is suppressing it. In statements to Entrepreneur on May 19, 2026, Laura Lorenzetti, VP and executive editor of LinkedIn, put it this way: "AI can truly help people unlock content creation. But it also means that many people can produce a lot of very low-quality content." The platform points to three things: generic posts written with AI without personal input, automation tools that generate comments, and clickbait videos with little substance.

The method they describe is relevant for understanding what can and cannot be "faked": human editors and content managers tag thousands of posts to determine what counts as generic and what as original, and with that, they train detection models. And the punishment is not deletion: it is limiting distribution, so that the post does not reach much beyond the direct network of the person who published it. This perfectly explains the most repeated symptom this year: "I get the same number of likes as always, from the same ten people, and nothing more".

The operational conclusion is not "don’t use AI." It is: the AI that serves you is the one that helps you structure and package what you know, not the one that fills the text field for you. A post that describes a real case from your work, with figures and the specific decision you made, does not resemble anything generated abstractly. A post that says "5 lessons about leadership I learned at the gym" resembles two hundred thousand.

Which format withstands the fall better: document posts

The PDF carousel —what LinkedIn calls a "document post"— is the format with the best ratio between reach and interaction in 2026. According to the study of format performance by AuthoredUp, based on over 3 million personal profile posts from March 2025 to February 2026, document posts achieve 1.39× the reach and 1.30× the interaction of the platform average. Metricool measures it absolutely and reaches the same conclusion: 1,451.19 impressions on average for carousels versus 605.57 for videos.

Comparison between a single slide post turned off and a five-slide carousel illuminated by a continuous light trail
Comparison between a single slide post turned off and a five-slide carousel illuminated by a continuous light trail
FormatReach (×average)Interaction (×average)% of posts
Survey1.78×0.37×1.16 %
Document / carousel1.39×1.30×4.88 %
Image1.20×1.33×57.16 %
Text only1.07×0.78×11.59 %
Video0.86×0.93×10.61 %
Article0.69×0.44×6.00 %
Reshare0.29×0.22×7.69 %

Source: AuthoredUp, 3 million personal profile posts, March 2025 – February 2026. Base 1.0× = platform average.

Three takeaways from that table that are worth more than the headline:

  1. The survey is a reach trap. 1.78× reach with 0.37× interaction means many people see it, but no one cares. In a ranking that measures whether what it teaches is annoying or not, that’s a liability.
  2. The document is the only format that wins in both columns at once. And it makes mechanical sense: a carousel pulls the person through several slides, so the time spent with your content is measured in tens of seconds, not in the second and a half it takes to read a paragraph.
  3. What almost no one does is what works best. Documents make up 4.88 % of posts. Single images, 57.16 %. Resharing, with a reach of 0.29×, accounts for 7.69 % of people’s effort.

Two more data points from Metricool to place the work in time: 50 % of total impressions come in the first 48 hours, and posts with a direct question receive 77.39 % more comments. If the carousel is what retains, the question in the first comment is what turns that retention into conversation.

And a warning for those who include links: on personal profiles, Metricool measures -27 % impressions and -20 % interactions when the post includes a link, whereas on company pages, the opposite happens (+51 % and +41 %). The same action, opposite result depending on where you publish from.

Your profile is already part of the ranking: the headline and the "About"

This is the consequence that almost no one has yet incorporated into their routine. If the Profile Relevance model decides whom to show your post to by comparing the topic of the post with the position, company, sector, and seniority of the reader, then your own profile is the document that the system reads to understand what you’re about. It’s no longer a business card that a recruiter looks at once a year: it’s an entry in the ranking.

Beam of light traversing the headline and description of a professional profile from which threads lead to various silhouettes of readers
Beam of light traversing the headline and description of a professional profile from which threads lead to various silhouettes of readers

The data that confirms this most strongly is the +17.7 % viral interaction in new members with limited history. When there’s no behavior to draw from, the system relies on what’s declared. And even if you’ve been on the platform for ten years, every new reader whom it wants to show you is in that situation regarding you: they have never read you, and the only thing to decide if they’re interested is the semantic fit between your post, your profile, and theirs.

What that means in concrete terms:

  • The headline has to say who you serve and for what, not what your internal title is. "Operations Director" doesn’t tell the model anything about who might be interested in what you write. "I reduce the logistical costs of food distributors" does.
  • The "About" is where the long context goes: the problem you address, with what type of client, with what results. It’s the longest and most declarative text you have, and now it gets read.
  • Your profile and your posts have to talk about the same topic. If you post about hiring and your profile says you do business development, you’re asking the system to resolve a contradiction, and it resolves it by not showing you.
  • Coherence over creativity. The goal is not a clever headline; it’s an unequivocal headline.

What to do this week, in order

If you had to focus on three things and nothing more, they would be these three, in this order. First the profile, because it conditions the distribution of everything you publish afterward; then the format, because it is the change with the greatest effect per unit of effort; and finally the routine, because 50% of impressions are made in 48 hours.

  1. Rewrite the headline and the "About" so that they say who you serve, in what sector, and with what result. Check that a stranger from your target audience would understand in five seconds if you are for them.
  2. Turn your best text post from the last six months into a PDF carousel. Do not invent new content: take what has already shown interest and give it the format that retains.
  3. Remove polls and reposts from the calendar. 1.78× reach with 0.37× interaction and 0.29× reach, respectively, do not justify the space.
  4. Replace "Do you agree?" with a real question that only someone in your profession can answer. Direct questions receive 77.39% more comments.
  5. Engage with the post for the first 48 hours. That’s when half of its total reach is decided, and comments with substance are what the ranking can distinguish from a "Great contribution!".

None of this is quick by hand. Turning an idea into a designed carousel, with a cover, readable slides, and a closing, is an afternoon of work if you do it in a design tool; and rewriting a headline until it is unambiguous takes fifteen versions. It is exactly the type of repetitive and formatted task where an AI tool contributes without replacing what you know.

Frequently Asked Questions

Why has my reach on LinkedIn decreased in 2026?

For three reasons at once. First, there is more competing content: LinkedIn reported a 14% year-on-year increase in content creation (Entrepreneur, May 19, 2026). Second, the ranking criteria have changed and now scores the fit between your post and the professional context of each reader, with the Profile Relevance model that the platform detailed on November 9, 2026. And third, the platform is actively limiting the distribution of generic AI content and engagement bait. In figures, Metricool measures a 10% drop in average impressions per post, from 924.42 to 831.96.

What is LinkedIn's Profile Relevance model?

It is the system that LinkedIn explained on November 9, 2026, in its engineering blog. It uses embeddings generated by an LLM to build a semantic summary of each member's professional context and measure how well each post fits with it, taking into account the topic, location, position, company, sector, seniority, and education. In LinkedIn's experiments, it reduced content hiding by 18.8% and increased daily active users by 1.8%.

Does LinkedIn use 360Brew to order the feed?

No. Tim Jurka, VP of engineering at LinkedIn, confirmed on April 14, 2026, that 360Brew was an internal test with a small group of members, who decided it did not fit the platform and closed the test. The original paper, published in 2025, was withdrawn from arXiv by its lead author. Any guide that explains "how to optimize for 360Brew" is optimizing for a model that is not in production.

Does LinkedIn penalize posts written with artificial intelligence?

It penalizes generic posts, not the use of the tool. Laura Lorenzetti, VP and executive editor of LinkedIn, explained in May 2026 that the platform targets three things: generic AI posts without original contribution, comments generated by automation, and substance-less engagement bait videos. Marked content is not deleted: its distribution is limited so that it does not go much beyond the author's direct network. A post with your data, your case, and your opinion does not fall into that category even if you wrote it with help.

What format has the most reach on LinkedIn in 2026?

The survey achieves the most gross reach (1.78× the average), but with disastrous interaction (0.37×). The format that wins in both areas is the document post, the PDF carousel: 1.39× reach and 1.30× interaction, according to AuthoredUp on more than 3 million posts between March 2025 and February 2026. Metricool confirms this absolutely, with 1,451.19 average impressions for carousels compared to 605.57 for video.

LinkedIn has not published any official figure, and the recommendations circulating (around 8-12 slides) come from commercial guides, not from the platform. What is grounded is the mechanism: the carousel works because it pulls the reader through several slides and extends the time they spend with your content. If each slide contributes something and the whole is read from start to finish, the exact amount is secondary.

Does my LinkedIn headline influence the reach of my posts?

Yes, and it is new. The Profile Relevance model compares the topic of your post with the position, company, sector, and seniority of each potential reader, so your profile is one of the inputs with which the system decides whom to show you. The clearest indication is that the improvement LinkedIn reports is much greater (+17.7%) among new members with little behavioral history: when there is no past activity, the system relies on what the profile states.

It depends on where you publish from, and the contrast is significant. Metricool measures that, on personal profiles, posts with links lose 27% of impressions and 20% of interactions; on company pages, however, they gain 51% and 41% respectively. The figures of "60% penalty for links" that you will see repeated have no official source from LinkedIn.

How long does it take for a post to stop growing?

The 2026 LinkedIn study by Metricool measures that 50% of total impressions occur in the first 48 hours. This makes the first two days the window where it is worth responding to comments and being present, and makes it less profitable to revisit a post a week later.

Where an AI tool fits into all this

The 2026 diagnosis is uncomfortable but actionable: reach is no longer gained by publishing more, but by being unequivocal about what you are about and using the formats that retain. And that translates into production work —a headline that leaves no doubts, an "About" that tells who you serve, well-designed carousels instead of loose paragraphs, posts that sustain reading— which is precisely what is repetitive. aiLinkedin, from AiToolsForApps, is a box of AI tools for covers, headlines, carousels, and LinkedIn posts designed for that part: solving the format and packaging so that you provide the criteria, data, and real case, which is exactly what the new ranking knows how to distinguish from generic text.

Set the criteria yourself and let the tool handle the format: generate your cover, your headline, your 'About', your PDF carousels, and your posts with aiLinkedin, no registration needed to try it out.

Set the criteria yourself and let the tool handle the format: generate your cover, your headline, your 'About', your PDF carousels, and your posts with aiLinkedin, no registration needed to try it out.

↑ Back to top