How To Adapt to LinkedIn's New Algorithm (Q2 2025 Breakdown)
By Nader Alnajjar
TLDR
Median impressions on LinkedIn are down 67% since 2023, but creators publishing niche, high-quality content are growing faster than before. The algorithm has not broken; it has got pickier about who it treats as an expert.
LinkedIn replaced its old mechanical ranking system with 360Brew, a 150-billion parameter AI model that reads your post, compares it to the expertise on your profile, and decides who should see it. Engagement pods are penalised, hooks carry 3-5x more weight, saves are now the most important metric.
Five shifts to act on: niche creators win on relevance, carousels (4.1x) and portrait visuals (86% better) beat text, AI-in-your-industry topics outperform broad leadership posts, posts now live 3-4 days so quality beats volume, and 2k-5k likes is the new breakout in a niche.
This week: audit the formats in your last 10 posts, pick one niche pillar with AI relevance, swap broad leadership content for storytelling and insight carousels, and make more of whatever your top 20% of posts did.
LinkedIn’s algorithm has changed, yet again… And we've broken down the data so you don't have to. These are the few takeaways that are most worth taking note of.
Build Leverage By Learning:
Why impressions are down (and why that's not necessarily a bad thing)
The 5 key algorithm shifts every creator needs to know
Examples of these shifts in practice
Practical next steps you can action this week
Why This Report Matters
If you’ve noticed your posts getting fewer views lately… you’re not imagining things.
Median impressions across LinkedIn are down 67% since 2023. Which is a pretty insane stat. However, on the other side of that coin, another interesting thing is happening: those who are creating more niche, high-quality content are actually growing faster than before.
So it’s not that the algorithm is ‘broken’. It’s just gotten pickier, valuing those who are clearly an expert in their field (an understandable reaction to the surge in AI slop on the platform).
We’ve been working closely with our partners at Saywhat and The Creator Accelerator to track these changes in real time. And in this newsletter, we’ll not only break down what they mean but also how these changes have manifested with real creator examples, and how you can start to implement them in your own content.
The Biggest Change: LinkedIn’s New AI Brain
Before we get into the data, you need to understand what's actually driving all of this.
LinkedIn scrapped its old ranking system entirely. They replaced it with a single AI model called 360Brew: a 150-billion parameter language model that actually reads and understands your content.
This is a pretty massive change.
The old system was mechanical. It tracked who you followed, what hashtags you used, what you clicked on. But the news system works differently. It infers relevance from meaning. It reads your post, compares it to what your profile says you're an expert in, and decides whether it makes sense to show it to a particular audience.
There are also a few other key changes:
Engagement pods are dead → AI can now detect coordinated engagement and penalise it.
Your profile matters more than ever → The algorithm checks whether your content matches your stated expertise.
Your hook carries more weight → The first 1-2 sentences of your post get 3-5x more processing attention from the AI.
Topical consistency is rewarded → Aim for 80%+ of your content to fall within 2-3 core topics over a 90-day window.
Saves are now the most important metric → We're seeing this internally at Lever too. A save tells the algorithm someone found your content worth returning to.
Comment depth beats comment volume → Ten thoughtful replies matter more than fifty "Great post!" comments.
Posting time matters less → The AI evaluates content quality now over when you publish
Links in comments don’t help anymore → If you have links, put them in the post itself, not the first comment.
TL;DR - LinkedIn's AI is smarter now. It rewards genuine expertise and penalises shortcuts. Which, honestly, is how it should've worked from the start.
5 Key Takeaways From The Algorithm Report
1️⃣ Impressions Are Down, But Niche Creators Are Coming Out On Top
Yes, overall reach has dropped. But creators with a defined voice, consistent format, and clear ICP (ideal client profile) are seeing more engagement than ever. The algorithm now prioritises relevance over raw reach.
What this means: If you're posting generic "leadership thoughts" to appeal to everyone, you're appealing to no one. The creators thriving right now are the ones who've picked a clear lane and stayed in it.
2️⃣ Format Priorities Have Shifted
The data here is clear. Carousels get 4.1x more reach than text posts. Infographics get 3.4x. And portrait images outperform landscape by 86%.
There’s also been a big shift toward story-led carousels as opposed to educational/value-driven ones.
What this means: Default to portrait visuals and story-led carousels. Quote graphics and random selfies aren't cutting it anymore unless there's genuine context behind them.
3️⃣ Topics Matter More Than Ever
AI, tech, and niche operational content are massively outperforming broad leadership posts. Even if AI isn't your industry, creators who talk about AI in their context are seeing success.
What this means: You don't need to become an AI influencer. But if you're a CFO, talk about AI in finance. If you're in recruitment, talk about AI in hiring. Meet the algorithm where it's paying attention.
4️⃣ Quality Over Quantity
Posts now perform 3-4 days instead of 24 hours. If your post is high-quality and thoughtful, it will have a longer life span regardless of how often you post.
What this means: Prioritise mid-funnel, high-value posts with strategic hooks. Don’t make any ‘filler’ content; make sure every post is well thought-through and created to a high standard.
5️⃣ Virality Has a New Definition
The days of 30k+, even 100k+ likes on every viral post are gone. In niche categories, 2k-5k likes now counts as breakout success, like the examples below.
What this means: Stop chasing big numbers. Chase resonance. Going deeper with fewer people who actually care about what you do is how you build a strong personal brand.
What This Looks Like in Practice
Some creators have already started adapting their content accordingly and are seeing success. If you want some inspiration on how to start, here are a few examples:
This post by Richard Harpin is a great example of how authenticity is being increasingly favoured: a personal story about some leadership lessons he’s learnt in his entrepreneurial journey. Even though the media is just a portrait picture of him, the post still got almost 4,000 likes. It shows how you can still deliver value through less rigid and more personal
On the AI front, we’ve seen creators who aren’t “AI founders” dominate by talking about AI within their niche. Take Paul Evans, for example. The main focus of his content is positioning for brands, but a he’s found a lot of success with exploring how AI intersects positioning development. Some examples include this post, and this one.
In terms of visuals, this carousel by Andrew Aziz shows how you can add a more personalised/story-based angle to what would otherwise be a purely value-driven carousel. Even if it means just re-thinking the title and adding a few pictures of you, it could 4x your post’s performance.
A few more points…
Do external links still kill reach? NO.
High-value link posts actually outperform now. The algorithm punishes low-effort link drops, not links themselves.
Do hashtags help? NO.
Using more than 3 hashtags leads to 70% worse performance. Just avoid them completely.
Should I post videos? YES (but only if you're in them).
Personal clips and demos outperform repurposed viral content. Faceless video compilations are of little to no value.
Is writing longer better? YES.
Posts between 1,250-3,000 characters with clear line spacing perform best. Give people something to actually get stuck into and find value in.
What You Can Do This Week
In terms of what these changes mean for you, and how you can start actioning them:
✅ Run a format audit → Look at your last 10 posts. How many used carousels, infographics, or portrait visuals? Adjust your content schedule so each of these are implemented once a week (at least).
✅ Pick a niche content pillar and go deep → Ideally one with AI relevance, even if it’s just how AI applies to your industry.
✅ Replace broad leadership content with storytelling and insight carousels → Prioritise genuine authenticity over preachy content.
✅ Track your outliers → Find your top 20% performing posts from the last 90 days. Make more content like those.
The algorithm’s going to keep evolving. That’s just how it works. But the fundamentals will pretty much always stay the same.
If you’re focused on creating genuinely useful content for a specific audience, I can almost guarantee you will do well on the platform.
Until next time,
Nader 🫡
P.S. If you’d rather have someone else worry about algorithm updates and produce content that stays ahead of the curve while you focus on running your business, reply to this email - let’s talk.
Otherwise, as always, reply to this email with any questions or just to say hi! It helps with deliverability and we genuinely read every response.
What this means for you
If you are a founder or executive building a founder-led brand, the shift to 360Brew works in your favour, provided you stop behaving like a general-interest creator. The model now reads your post and checks it against what your profile says you are an expert in, so a CEO whose feed is 80% about one or two subjects over a 90-day window is exactly what it is built to reward. The 67% fall in median impressions is mostly felt by people posting broad leadership thoughts to everyone; the creators Nader points to, such as Richard Harpin getting close to 4,000 likes on a personal leadership story with nothing more than a portrait photo, are winning on relevance and resonance rather than reach. Chasing 30k-like virality is no longer the game. In a defined niche, 2k to 5k likes from the right people is breakout, and those are the people who become clients, investors and hires.
The practical shifts are mechanical enough to fix in a week. Make your profile a precise statement of the expertise you want to be shown for, because the algorithm cross-references it. Spend real effort on the first two sentences of every post, since the hook receives 3 to 5 times more processing attention. Move your visuals to portrait and your carousels to story-led, drop hashtags, put links in the post rather than the first comment, and treat saves and thoughtful replies as the metrics that matter. Then find an angle on AI inside your own industry, because that is where the model is paying attention, and audit your last 90 days for the top 20% of posts so you can make more of what already works.
Frequently asked questions
Why are LinkedIn impressions down and does it matter?
Median impressions across LinkedIn are down 67% since 2023, so if your posts are getting fewer views you are not imagining it. Nader's point is that this is not the algorithm breaking. It has got pickier, favouring people who are clearly experts in their field, partly as a reaction to the surge in AI slop on the platform. On the other side of that drop, creators producing more niche, high-quality content are growing faster than before. Overall reach is lower, but creators with a defined voice, consistent format and a clear ideal client profile are seeing more engagement than ever, because relevance now beats raw reach.
What is 360Brew and how does LinkedIn's new algorithm work?
LinkedIn scrapped its old ranking system entirely and replaced it with a single AI model called 360Brew, a 150-billion parameter language model that reads and understands your content. The old system was mechanical: it tracked who you followed, which hashtags you used and what you clicked. The new one infers relevance from meaning. It reads your post, compares it to what your profile says you are an expert in, and decides whether it makes sense to show it to a particular audience. It can also detect coordinated engagement, so engagement pods are penalised, and it gives the first one or two sentences of a post 3 to 5 times more processing attention.
Which post formats perform best on LinkedIn now?
The data in the report is clear. Carousels get 4.1 times more reach than text posts and infographics get 3.4 times, while portrait images outperform landscape by 86%. There has also been a big shift toward story-led carousels rather than purely educational or value-driven ones, so Nader's advice is to default to portrait visuals and story-led carousels; quote graphics and random selfies no longer cut it without genuine context. Video works if you are in it, since personal clips and demos outperform repurposed viral content. Longer posts of 1,250 to 3,000 characters with clear line spacing perform best, and more than three hashtags leads to 70% worse performance.
Do external links and hashtags still hurt LinkedIn reach?
External links no longer kill reach. High-value link posts actually outperform now, because the algorithm punishes low-effort link drops rather than links themselves. One change to note: links in the first comment no longer help, so if you have a link, put it in the post itself. Hashtags are a different story. Using more than three leads to 70% worse performance, and the newsletter's advice is to avoid them completely. Posting time also matters less than it did, since the AI evaluates content quality over when you publish, and a good post now performs for 3 to 4 days instead of 24 hours.
Related reads
About the author
Nader Alnajjar is co-founder of LeverBrands, where he builds personal brands and content ecosystems for founders and executives. More at leverbrands.com/about and on LinkedIn.