LinkedIn algorithm 2026: how the feed really works, according to LinkedIn's engineers

What changed in the LinkedIn algorithm in 2026?

On 12 March 2026, LinkedIn's engineering team published a detailed post on the next generation of its feed. It is the best source there is, because it is not a creator's guess but a description of the system by the people who built it. According to the post, the feed serves more than 1.3 billion professionals and is moving to a new ranking system powered by large language models that "better understands what a post is actually about and how it relates to a member's evolving interests and career goals".

There are two big changes:

1. One retrieval system instead of many. Until now the feed gathered posts from several separate sources: your network's activity in chronological order, trending posts in your region, "people like you liked this", and more. Now a language model turns every post and every member into a representation of meaning and picks posts by topical closeness, even when the words differ. LinkedIn writes that the result is content "from authors you don't follow, on topics you care about".
2. Ranking that reads the story, not the moment. The new ranking model (a Generative Recommender) looks at more than a thousand of each member's past actions as a sequence, like a professional journey, instead of predicting each impression on its own.

There is research underneath too: LinkedIn published a paper on 360Brew, a foundation model for ranking and recommendations across its products. But what runs in the feed today is described in the March post, and this guide is built on it.

How does LinkedIn decide who sees your post?

According to LinkedIn's post, the system weighs two kinds of information about every reader:

• Their profile: industry, experience, skills and location.
• Their behaviour over time: what they read, liked, commented on, returned to, or simply scrolled past.

For every post, the model gets the format, the author's details (name, headline, company and industry), how much engagement it has compared with other posts, and the text itself. The actions the model learns from explicitly include "long dwells", likes, comments and shares, and skips too.

Three practical conclusions, in my reading:

1. Your headline is part of the post. The model reads the author's headline and company. A clear headline about what you do helps it understand who the post is for.
2. A consistent topic helps the model. It matches posts to readers by meaning. Someone who writes about the same field again and again gives it a clear picture. Someone who jumps between ten topics, less so.
3. A like isn't everything. Scrolling past is a negative signal and a long dwell is a positive one. A post people read to the end is worth more than one they like on the way past.

LinkedIn adds in its Help Center that its algorithms also "filter out or taper distribution" of low-quality or unsafe content, and its engineers say they audit that posts from different creators "compete on equal footing". Nothing in either source promises reach for follower count.

Why did my LinkedIn reach drop?

Probably not because you did something wrong. AuthoredUp analysed more than three million posts (March 2025 to February 2026, against the year before) and found two trends:

• Big accounts lost half their reach. For the top 5% of profiles, median impressions per post fell from 13,711 to 6,868. For all profiles the median stayed stable, at about 860.
• Engagement is falling across the platform. Median engagement per post dropped from 29 to 25, and for the top 5% from 198 to 111. The likely reason: more people are posting, so there is more competition for the same feed.

That fits what LinkedIn's engineers describe. When ranking is done per reader, by what is relevant to them, a big audience no longer guarantees big reach. Every post has to earn its place again, which is good news for smaller creators.

What about company pages? In the same analysis, a personal-profile post has a median of 1,006 impressions and a company-page post 659. The difference between the two, and what a company page is still good for, is on the LinkedIn management page.

Which type of LinkedIn post works best in 2026?

According to AuthoredUp, every format lost reach over the past year, but not equally. A multiplier of 1.39 means 1.39 times the median reach:

Two things stand out. First, polls get the most reach but almost no comments: people vote and move on. Second, video, the format LinkedIn pushed in 2024, fell below average. Documents hold up best, and AuthoredUp's explanation makes sense: every swipe is a few more seconds of reading, which is exactly the "long dwell" signal LinkedIn describes.

But don't turn everything into a carousel. We saw it in the analysis of 702 posts by CEOs on LinkedIn Top Startups: what built their voice was the content, not the format. A good format amplifies a good post. It doesn't rescue an empty one.

What is the best time to post on LinkedIn?

For a personal profile, a weekday morning: between 8:00 and 11:00 in your audience's local time, with the most common peak around 9:00, according to AuthoredUp's analysis of more than three million posts (updated October 2026). Two other large studies published in 2026 found later hours, because they measured different accounts:

Where all three agree: weekdays beat weekends. Buffer reports that weekend engagement drops off significantly, Sprout Social found no optimal times on Saturday or Sunday, and for company pages AuthoredUp measured weekdays about 48% above weekends.

Why do the studies disagree on the hour?

Because they count different things. Buffer looks at posts scheduled through its own tool, without separating personal profiles from company pages. Sprout Social counts engagement on its customers' profiles across six networks, not LinkedIn alone. AuthoredUp analyses LinkedIn only and reports personal profiles and company pages separately. For a CEO or founder posting from a personal profile, AuthoredUp is the closest match, which is why the morning answer leads here. Buffer adds a fair warning of its own: strong engagement doesn't automatically mean more reach.

What is the best day to post on LinkedIn?

For a personal profile, almost any working day: in AuthoredUp's data the gap between the best and worst day is under 10%. For a company page, the middle of the week: a US company page's Wednesday runs about 1.7 times higher than its Sunday. Buffer and Sprout Social also put Tuesday to Thursday on top, and Buffer found Monday the weakest weekday.

What is the best time to post on LinkedIn in Israel?

None of the three studies breaks out Israel. Since all of them work in the audience's local time, the logic carries over: for an Israeli audience, Sunday to Thursday mornings (Friday and Saturday are the weekend). For an American audience, their morning, which is afternoon or evening in Israel.

Most important: AuthoredUp found that what you post moves engagement far more than when you post, and that the gap between formats dwarfs the gap between the best and worst hour. A good post at a mediocre hour beats a mediocre post at the perfect time. To find your own best time, pick two slots the studies support, say 9:00 and 16:00 on working days, post in each for a month, and compare impressions per post in your LinkedIn analytics. Your own audience beats any average.

What about the "golden hour"?

LinkedIn hasn't published a "the first hour decides everything" rule. What it has written: when a post is published or gains engagement, the system updates its representation within minutes. And in July 2025 its VP of product, Gyanda Sachdeva, told Business Insider that LinkedIn was testing how recent content needs to be to stay relevant, after members saw posts two or three weeks old in their feeds. So a good post can keep working long after the first hour. Early comments help, but they are not a verdict.

Do hashtags still work on LinkedIn?

Not like they used to. According to Sprout Social, LinkedIn removed hashtag following, took hashtags off profile displays and stopped suggesting them in search. A hashtag no longer puts your post in the feed of people who follow it, because those followers no longer exist.

What's left: a hashtag is one more word in the text the model reads, and it can help in search. My recommendation: one to three relevant hashtags at the end, or none. And never in the first line, which is where the sentence that makes people stop belongs.

Do links in a post hurt reach?

By AuthoredUp's estimate, based on its earlier research and Richard van der Blom's report, a personal-profile post with an external link in the body gets 40% to 50% less reach. It is an estimate, not an official LinkedIn number, but the direction is consistent: LinkedIn wants people to stay on the platform.

What to do: write a post that delivers its full value on LinkedIn, and put the link in the first comment for anyone who wants more. If the post only works when someone clicks the link, most readers will miss it anyway.

How often should you post on LinkedIn?

According to AuthoredUp, four to five posts a week give the highest engagement rate (2.60%) and 28% more impressions per post than posting once a week. Above eight posts a week, both engagement (1.79%) and reach per post drop sharply.

For the CEOs I write for it is two to three posts a week, because for most of them that is the pace they can sustain without dropping quality. That is the point: consistency over months beats a one-week burst. It usually takes 3 to 6 months of consistent presence to get meaningful organic leads. To keep that pace without starting from zero every time, I use the content waterfall: one central idea that breaks into weeks of posts.

Does LinkedIn penalise AI-written posts?

In AuthoredUp's 2025 data, AI-generated posts got 30% less reach and 55% less engagement than human-written ones. LinkedIn hasn't published such a rule itself, and there is no way to tell whether it is an algorithmic penalty or simply readers scrolling past when something sounds generic. For the outcome it doesn't matter: scrolling past is a negative signal.

That is also why, at the agency, I write managed clients' posts by hand, in their voice. The stories, numbers and opinions come from them, and that is what makes someone stop and read. If you write yourself, the LinkedIn text formatter shows how your post will look and where LinkedIn cuts it.

What does this mean for CEOs?

The new algorithm looks for a match between a post and a reader, not for viral posts. For a CEO that means one thing: you don't need 100,000 views. You need the right 100 decision-makers. A post I wrote for a CEO got only 20 likes and brought in ₪50,000 within 8 hours.

1. Pick three or four topics you want to be known for, and write about them again and again. That is how the model learns who to show you to.
2. Write for one specific reader, the client you want to meet, not for "everyone".
3. Deliver value inside the post: a story, a number from your work, a decision you made. Something worth reading to the end and saving.
4. Reply to comments and comment on the posts of the people you want to meet. That puts you in the history the model reads for them.
5. Don't measure likes. Measure conversations: inquiries, meetings and messages. Content alone isn't always enough for that, and LinkedIn outreach completes it.

If you don't have time to do it yourself, that is exactly what LinkedIn ghostwriting is: writing in your voice, on a fixed calendar, after a strategy. And one more thing worth remembering: LinkedIn content is also cited in ChatGPT and Google answers, which I cover in the guide on how to get cited by ChatGPT.

The bottom line

There is no trick to the LinkedIn algorithm in 2026, and according to its engineers that is the point. The system reads the post, knows the readers, and checks whether they stay. What works is what always worked, minus the shortcuts: a clear topic, real value, a human voice and consistency.