Why is My LinkedIn Reach Dropping? The 360Brew Algorithm Explained (And How to Fix Your LinkedIn Content Strategy)

¶Why have my LinkedIn posts stopped getting views?

If your LinkedIn posts have felt like they’re shouting into the void recently, I have good news and bad news.

The good news: it’s not just you.

The bad news: it’s basically everyone.

In January 2025, LinkedIn published a research paper describing a brand new AI model called 360Brew. It’s a 150-billion-parameter system designed to completely overhaul how content gets distributed on the platform.

Then, in a move that tells you everything about how significant this is, they withdrew the paper a few months later because it apparently wasn’t supposed to be public yet.

Oops.

But the cat was well and truly out of the bag. (Or, the lime was out of the bowl, if you prefer). And since then, the effects have been impossible to ignore. LinkedIn engagement is down across the board. LinkedIn impressions have dropped for almost everyone. And the old tricks - hashtags, pods, posting five times a day - don’t work anymore.

So we’ve done what we always do when something confuses us: We went down a massive research rabbit hole. And we’re going to try to explain what we found in a way that’s actually useful (and doesn’t require a computer science degree).

Grab a cuppa. This is a big one.

¶What is LinkedIn’s 360Brew algorithm?

OK, here’s the short version. 360Brew is LinkedIn’s new AI brain. It’s a decoder-only foundation model (think: the same type of technology that powers ChatGPT) that was built to replace thousands of separate, specialised ranking systems with one unified engine that can read, understand, and rank content across LinkedIn’s entire platform. (If you want the long and complicated version, here’s the full research paper).

The name is a clue: a 360-degree view of LinkedIn’s ecosystem, “brewed” into one model. (LinkedIn engineers love a pun, and we’re very pleased to hear it!)

The paper was published on 27 January 2025 by 23 authors from LinkedIn’s Foundation AI Technologies team. The model was built on top of Mixtral 8x22, a Mixture of Experts architecture, and trained on LinkedIn’s own data.

Here’s the fundamental shift, and it’s a big one:

The old LinkedIn algorithm counted things. How many likes in the first hour? How many hashtags did you use? How quickly did people comment? Engineers hand-built thousands of these numerical signals and fed them into separate little models. (LinkedIn’s previous LiRank system)

The new LinkedIn algorithm reads and understands your content. It builds a text prompt describing the full context - who you are, what you’ve been engaging with, what the post is about - and uses AI to predict whether you’ll find it valuable.

It’s the difference between someone counting words on a page and someone actually reading the book.

One model now handles 30+ different ranking tasks across at least eight LinkedIn surfaces: your feed, job recommendations, People You May Know, ads, search, notifications, and more.

¶Is 360Brew running LinkedIn’s algorithm right now?

This is where it gets a bit murky. And where most of the commentary online gets it wrong.
LinkedIn has never officially confirmed that the specific 360Brew model from the research paper is directly powering the feed.

What they did announce - in a March 2026 engineering blog post by Senior Staff TPM Hristo Danchev - is a new two-stage system:

• Stage 1: An LLM-powered retrieval system that narrows millions of candidate posts down to roughly 2,000 using AI-generated embeddings and cosine similarity matching.
• Stage 2: A Generative Recommender that processes over 1,000 of your recent interactions as a chronological sequence to rank those 2,000 posts in order of predicted relevance to you.

The production system draws heavily from 360Brew’s innovations but isn’t identical to what the paper described. Interestingly, one technical report notes that a section of LinkedIn’s internal specs is titled “The LLM-Ranker Was Evaluated and Rejected” - meaning they tested using 360Brew as a direct ranker and decided against it for the production feed.

But here’s what matters for us: “360Brew” has become shorthand for the whole LinkedIn algorithm overhaul. Whether or not the exact research model is running the show, the signals, strategies, and content principles are the same. So that’s what we’re going to focus on.

¶How much has LinkedIn reach dropped? (Spoiler: a lot)

Let’s look at the actual numbers. Because when I say “your LinkedIn reach has probably dropped,” I’m not guessing.

Analysis of over 3 million LinkedIn posts and Richard van der Blom’s Algorithm Insights Report (analysing 1.8 million posts) paint a consistent picture:

Sources for table data: Agorapulse LinkedIn Algorithm 2026 | UpGrowth 360Brew Report | Pettauer Visibility Analysis

98%. Ninety-eight percent of LinkedIn users saw their reach decline. So if you’ve been asking yourself “why is nobody seeing my LinkedIn posts?”, it’s not your content. It’s the algorithm.

But (and this is the interesting bit) the reach didn’t disappear. It concentrated.

¶The LinkedIn reach gap is getting bigger

The top 1% of LinkedIn posts now outperform the median by 237x. Two hundred and thirty-seven times. That’s not a typo.

Source: UpGrowth 360Brew analysis

Only 1.1% of LinkedIn’s user base posts weekly. Just 7.1% have posted at all in the last three months. LinkedIn is rewarding the people who show up consistently with genuinely good content - and taking reach away from everyone else.

The message is clear: one brilliant post is now worth more than ten mediocre ones. By a lot.

¶How LinkedIn engagement works now: the new rules

OK, this is the bit that matters most. Because the entire hierarchy of what makes a LinkedIn post perform well has been flipped upside down.

¶LinkedIn saves are now the most important engagement signal

Under the old algorithm, likes were king. Get enough quick likes, your post would spread. Now? One save drives approximately 5x more reach than a like and 2x more than a comment.

Let that sink in. If 10 people save your post, the algorithm treats it roughly the same as 50 people liking it. LinkedIn is now explicitly optimising for content people want to come back to - not content they passively scroll past and double-tap.

This completely changes what you should be creating. More on that in a sec.

¶LinkedIn dwell time is now a primary ranking signal

How long someone spends reading your post is now a primary signal. If people scroll past quickly, the algorithm takes that as a negative signal. If they stop, read the whole thing, maybe expand it - that’s a strong positive.

This is why longer, meatier posts tend to outperform quick hot takes now. The algorithm is literally timing how long people spend with your content.

¶LinkedIn comment quality now beats comment quantity

Not all comments are equal anymore. A thread where three or more people have a proper back-and-forth conversation triggers 5.2x more reach. A single “Great post! 🚀”? Basically worthless. And if too many of those appear suspiciously quickly, it might actually flag you for using engagement pods.

(More on that lovely topic shortly.)

¶LinkedIn posts with delayed engagement perform better

Posts that receive saves and comments 24-72 hours after publishing perform 4-6x better in suggested feeds. This extends the content’s lifespan well beyond the old “first hour” window that used to determine everything.

LinkedIn VP Gyanda Sachdeva even confirmed that the platform intentionally shows posts up to 3 weeks old at the top of feeds. Your LinkedIn content now has a much longer shelf life than it used to. Which is brilliant news if your content is actually good.

¶LinkedIn likes are now the weakest engagement signal

Not useless. But the lowest weight of any engagement type. Which is quite the plot twist for a platform that built its entire engagement culture around the like button.

¶Do LinkedIn hashtags still work? No. They’re dead.

We’ve had a feeling about this for a while. And now it’s confirmed. LinkedIn VP Sachdeva said it herself: hashtags “do not impact distribution.”

Full stop. All those years of advice about using 3–5 hashtags on every LinkedIn post? It was doing absolutely nothing. We aren’t going to pretend we’re not a tiny bit smug about this.

So if you’re still carefully selecting your LinkedIn hashtags before every post… stop. Use that time to write a better opening line instead.

¶What does the LinkedIn algorithm penalise now?

Some of this might sting a little. Sorry in advance.

¶LinkedIn pods are being actively destroyed (yay!)

VP Sachdeva stated in November 2025: “Our goal is to make engagement pods entirely ineffective.”

LinkedIn now maps what it calls “Coordinated Activity Rings” - clusters of users who always engage with each other within minutes of posting. If you’re flagged, you get a shadow ban with a 60-90 day recovery period. As of February 2026, comments posted through third-party scripts and browser plugins are removed from the “Most Relevant” section.

Lempod, one of the biggest pod tools, has been banned and removed from the Chrome Web Store.

We’ve been banging on about pods for years, so we’re not going to pretend we aren’t pleased. But we get why people are tempted to use them. If you’re still in one though - now is the time to leave. The 60-90 day recovery clock starts when you stop.

¶LinkedIn engagement bait is being suppressed

“Comment YES if you agree!” “Like if you’ve been there!” “Tag someone who needs to see this!”

All detected. All suppressed. The LinkedIn algorithm can now identify these patterns and reduce your reach accordingly.

This includes “DM me KEYWORD” CTAs, by the way. They’re engagement-bait-adjacent. I know, I know - everyone does them. We’ve used them. But the algorithm isn’t a fan.

¶AI-generated LinkedIn content is being deprioritised

The algorithm can now tell the difference between your actual voice and generic ChatGPT slop. Lexical diversity and specificity are used as authenticity markers. If your LinkedIn posts sound like they were written by a robot on its lunch break, the algorithm notices.

(I say this as someone who uses AI all the time. The key is using it as a tool to enhance your ideas, not as a replacement for having ideas in the first place.)

¶Off-topic LinkedIn posts get penalised (yes, really)

This is the one that caught me off guard. The algorithm now cross-references your post content against your LinkedIn profile - your headline, About section, experience, skills - to check whether you’re posting about things you’re actually an expert in. Practitioners call it the “Expertise Mismatch Penalty.”

So if your LinkedIn headline says “LinkedIn Strategist” and you post about your holiday in Egypt, the algorithm sees a mismatch. (Guilty!) It doesn’t know it’s a lovely personal story your friends will enjoy. It just sees a LinkedIn person posting about pyramids and doesn’t know who to show it to.

(Yes, I’m speaking from experience. My Egypt posts did not do well.)

You don’t have to be a robot who never posts anything personal. But it does mean the 80/20 rule matters more than ever: 80% of your content should be within your 2–3 core topic areas, and the other 20% should ideally tie back in somehow.

¶How does the LinkedIn algorithm decide who sees your posts?

Understanding the mechanics helps explain why some LinkedIn posts fly and others flop. Here’s a simplified version of what happens when you hit “Post”:

1. Retrieval. Your post enters a pool with millions of others. The LLM-powered system reads your text, compares it to member profiles and recent behaviour, and narrows the pool to about 2,000 candidates for each user’s feed.
2. Ranking. A Generative Recommender looks at over 1,000 of each user’s recent interactions to rank those 2,000 posts. It predicts both passive engagement (will they actually read it?) and active engagement (will they like, comment, share, or save it?).
3. Profile matching. Your LinkedIn profile is processed to create a semantic fingerprint of your expertise. This is matched against potential viewers’ interests. Better match = more likely to appear in their feed.
4. Engagement feedback. Early signals then determine whether distribution expands or contracts. Saves, dwell time, and substantive comments trigger wider distribution. Quick scroll-pasts do the opposite.

One clever engineering detail: the system uses percentile bucketing instead of raw numbers. So a post with 50 views from a 500-connection account is treated differently to a post with 50 views from a 50,000-follower account. This delivered a 30x improvement in how the system matches content to relevance, and a 15% improvement in accuracy.

Which means small accounts with engaged audiences aren’t disadvantaged against massive accounts with lazy ones. Nice.

¶Which LinkedIn content formats work best in 2026?

The algorithm is technically format-agnostic - it cares about content quality, not whether it’s a carousel or a text post. But in practice, performance varies massively because of how people interact with each format.

The external links information is interesting. The old advice was always “never put links in your LinkedIn posts.” That seems to have shifted - slightly. Still test it with your own audience before going wild, but it’s interesting.

¶How to beat the LinkedIn algorithm in 2026 (OK, not “beat” - work with)

Here’s where we get practical. Based on everything above, here’s what our new LinkedIn content strategy playbook looks like.

¶1. Your LinkedIn profile is now your most important piece of content

The algorithm performs what practitioners call a “Profile-Content Audition” - it cross-references every post you publish against your headline, About section, experience, and skills.

If your profile is vague, clever-but-unclear, or doesn’t mention your actual expertise, the algorithm can’t figure out who to show your content to. Your profile needs to clearly and plainly state your expertise - clever wordplay and vague taglines are now counterproductive.

It takes approximately 90 days of consistent, niche-aligned posting for the system to fully categorise your expertise domain. So the sooner you tighten your profile, the sooner the algorithm starts working for you instead of against you.

¶2. Create LinkedIn content people want to save (not just like)

Since saves are worth 5x likes, the question to ask before every post is: “Would someone bookmark this to come back to later?”

The types of LinkedIn content that get saved: frameworks, data breakdowns, step-by-step guides, templates, checklists, contrarian takes backed by evidence, and anything someone would screenshot or send to a colleague.

Personal stories and hot takes can still work. But under this algorithm, they’re on hard mode because they don’t typically trigger the save behaviour.

¶3. Write better LinkedIn hooks

The first 1-2 sentences of your LinkedIn post carry disproportionate weight with the algorithm. Vague, generic openings get your post deprioritised before a human ever sees it.

• Boring: “I’ve been thinking about productivity lately.”
• Better: “I tracked every minute of my working day for 6 weeks. Here’s what I found.”

Names. Numbers. Specifics. Write your opening line like a journalist - lead with the hook, save the throat-clearing for never.

¶4. Get real conversations going under your LinkedIn posts

Multi-person comment threads with genuine back-and-forth are one of the strongest signals. So instead of ending your post with “What do you think?” (too vague), ask something specific enough that people have genuinely different answers. And reply to every comment with something substantive - the algorithm is watching thread depth, not just comment count.

¶5. Post less on LinkedIn, but make each post better

The optimal LinkedIn posting frequency is 2-5 times per week. Optimal post length is 800-1,000 characters.

We tested this ourselves. I went from 3 posts a week to 5 on my personal profile for about 6 weeks. Impressions barely changed. Sales didn’t change at all. So I went back to 3. Quality over quantity has never been more true on LinkedIn.

¶6. Stay in your lane on LinkedIn (mostly)

Aim for 80% of your LinkedIn content to be within your 2-3 core topic areas. The algorithm needs consistency to build your topical authority.

Personal posts aren’t banned. But “I went on holiday” is off-topic. “I went on holiday and here’s what it taught me about switching off as a founder” ties back in. See the difference?

¶7. Stop using engagement bait CTAs on LinkedIn

“DM me KEYWORD”, “Comment YES to receive”, “Like and repost to enter”. The algorithm detects and suppresses these. Replace them with direct links, genuine questions, or natural conversation starters.

¶8. Use LinkedIn carousels for your best content

Carousels still have the highest reach multiplier because they create high dwell time (all that swiping) and the text-rich format lets the algorithm fully index the content. If you’ve got a framework, a how-to, or a data-driven insight, a carousel is probably the best format for it right now.

¶What is LinkedIn trying to become?

When you zoom out from the tactical stuff, the picture is pretty clear. LinkedIn is deliberately repositioning itself as a professional knowledge-sharing platform, not a viral content engine. (I think we have enough of those!)

As the March 2026 engineering announcement put it, the system is designed to understand “what a post is actually about and how it relates to a member’s evolving interests and career goals.”

That 47% median reach decline? Not a bug. A feature. A deliberate reallocation of attention from noise to signal.

Which, if you think about it, is actually great news for those of us who have real expertise and are willing to share it generously. The playing field is being levelled against the pod users, the engagement baiters, and the people who’ve been gaming LinkedIn for years.

For the first time in ages, being genuine and being strategic can be the same thing.

And that makes me happy.

¶The era of gaming LinkedIn is over. The era of earning it has begun

I know that sounds dramatic. But the numbers back it up.

Real expertise, real authenticity, and content worth saving now determine who gets seen on LinkedIn. Engagement hacks, pods, and off-topic posting are being actively suppressed.

Your LinkedIn profile matters more than ever. And the creators who align their profiles, specialise their topics, and write content worth bookmarking will find the algorithm rewards them with longer content lifespans and more targeted distribution.

Those who don’t? Increasingly quiet feeds.

The good news? You’re reading this, which means you’re already ahead of the 98% of people who haven’t figured out what changed yet.

Now go update your LinkedIn profile. I’ll wait.