Did your LinkedIn impressions drop?
Here’s what actually changed.
If your posts used to hit 5,000 impressions and now stop at 800, you are not alone.
Multiple analyses across millions of posts show that LinkedIn reach has dropped roughly 40–50 percent in the last year for most creators.
This is not a bug.
LinkedIn rebuilt the whole machine under the hood.
What LinkedIn actually changed
Until 2025, LinkedIn used a stack of different models: one for feed, one for jobs, one for “people you may know”, and so on.
Now there is one big brain: 360Brew, a 150B-parameter AI model trained only on LinkedIn data.
(paper here)
Instead of asking only “which post gets the most quick reactions?”, 360Brew reads your profile, your post, and the reader’s context together, and asks a different question:
“Is this the right person, saying the right thing, to the right audience?”
If the answer is “I’m not sure”, your post simply does not travel very far.
From “who you know” to “what you talk about”
The old LinkedIn was about the social graph: your network, your connections, your follower count.
The new one is closer to an interest graph:
What you talk about
What people like you read and save
What kind of conversations your posts start
This is why people with 5,000 followers can suddenly outperform people with 50,000 followers: the system prefers relevance and depth of engagement over raw audience size.
Your profile is now part of the ranking pipeline, not just a static page. Misaligned profile and content = weak relevance score = low reach, no matter how good the writing looks on its own.
Why your reach really dropped
Let’s cut the drama and list the main real reasons, based on data from AuthoredUp, and multiple LinkedIn analytics reports.
1. The test‑audience spiral
Every post starts with a small test group from your network. If they ignore it, the post dies there.
When you hit a bad streak, something sneaky happens:
Post underperforms
Next test audience is slightly worse
Next post underperforms again
Audience quality degrades further
Some data shows that in 67 percent of “my reach suddenly died” cases, impressions recovered in 2–3 weeks with no big changes, which means it was a spiral, not a permanent punishment.
2. Profile and content do not match
360Brew ranks profile + content together.
If your headline says “Backend Engineer” and your last 15 posts are generic career advice, morning routines, and soft skills, the system has no clear idea what you are an expert in.
Low clarity = low relevance = low reach.
You are basically shipping a microservice with no clear API.
The router does not know when to send traffic to you, so it just… does not.
3. Generic AI / templated content
LinkedIn’s own explanations and third‑party analyses all point in the same direction: the feed is full of AI‑generated sludge, and 360Brew is being trained to detect it.
Signals that correlate with underperformance:
Same structure on every post (identical hook length, three equal paragraphs, question at the end)
Recycled “viral” templates with zero personal detail
Phrases that look like they came from a copy‑paste prompt, not from a real story
It turns out that posts with highly repetitive structure saw sudden, lasting drops in distribution, even when the topics were good.
4. Fake engagement and pods
Engagement pods and “comment for comment” groups are now easy to detect:
Same 20 people, always commenting in the first 5–10 minutes
Same short comments, with no depth
Cross‑posting patterns across accounts
On average, creators who were detected as part of pods saw their organic (non‑pod) engagement drop by 30–45 percent for 4–8 weeks.
LinkedIn is not just ignoring fake engagement; it is using it as a negative signal.
5. Outbound links and over‑posting
Several large‑scale analyses show two patterns in 2025–2026:
Posts with outbound links often get 50–70 percent less reach on average
Posting every single day, especially several times per day, correlates with worse performance per post
Is this “official”? No.
Is it visible in the data? Yes — enough that most growth experts now tell people to go easy on links and avoid trying to brute‑force reach with volume.
What 360Brew actually rewards
Now the part you care about: what to aim at instead.
Across AuthoredUp’s millions of posts and multiple independent breakdowns, three signals show up again and again.
1. Saves
Saves are the strongest signal you can realistically influence.
AuthoredUp data: saves drive around 5x more reach than a like, and around 2x more than a simple comment.
Saved posts stay in feeds longer and get resurfaced more often, because they look like “reference material”.
If nobody ever saves your posts, you are basically writing tweets in a place that wants documentation.
2. Dwell time
Dwell time = how long people actually stay on your post.
One study shows posts with 61+ seconds of dwell time have engagement rates more than 10x higher than posts that get only 0–3 seconds.
Multiple algorithm breakdowns say the same thing: a post with fewer likes but strong dwell time will beat a post with many likes that everyone scrolls past.
Carousels and longer text posts win here because they inherently take time to consume.
3. Long, relevant comments
360Brew can read language. It knows the difference between “great post!” and a three‑sentence answer with a concrete example.
Posts that spark real conversations stay visible longer.
Comments from people who look credible in your niche are worth much more than random low‑relevance profiles.
Likes are the weakest of the important signals. Good to have, not enough by themselves.
The engineer’s checklist
Before blaming the platform, run this basic debug pass on your account:
1. Profile → Do headline and About clearly match the topics you post about?
2. Last 20 posts → Are they about 2–3 tight themes, or all over the place?
3. Structure → Do your posts all look and “sound” the same?
4. Engagement → Are you getting real comments, or mostly quick likes?
5. Behavior → Any automation tools, pods, or growth hacks in the last 3 months?
6. Links → How many posts in the last 10 had outbound links?Then compare with this:
Reach dropped 40–60 percent but fluctuates from post to post → normal in 2026, most people see it.
Reach dropped 80 percent and stayed there for 10+ posts → you probably triggered one of the five patterns above.
In 4 percent of cases there was a true, sustained distribution hit, but every single one had at least one clear trigger: pods, links spam, guideline issues, or automation.
Stop guessing. Start shipping.
Here is the practical part — what to actually do for the next 90 days.
1. Lock in your topics and profile
Pick 2–3 topics that are close to your real work. For example:
“Backend performance and scaling stories”
“How we ship features faster without breaking prod”
“Lessons learned mentoring junior devs”
Update your:
Headline — make the topics obvious
About — 3–5 lines that say what you build, who you help, and what you write about
Featured — pin 2–3 posts or carousels that match those themes
You are telling 360Brew “this is my niche”.
Give it a consistent signal for three months.
2. Design posts for saves and dwell time
When you write, don’t think “how do I get likes?”.
Think “how do I make this worth saving?”.
Good candidates:
Step‑by‑step breakdown of how you solved a production incident
A small framework you actually use at work (how you estimate tasks, how you debug, how you do code reviews)
A short “field guide” carousel: “How we reduced cold start by 40 percent”, “How I onboard a new dev in 7 days”
Aim for:
200–400 words of clear text, broken into short paragraphs
Real numbers, real trade‑offs, concrete decisions
One main point per post, not a shopping list of tips
If someone on your team says “I will save this to show it to a junior dev”, you are on the right track.
3. Write for comments, not reactions
At the end of a post, don’t ask “what do you think?”.
Ask something that an engineer can answer quickly from experience:
“What is the worst rollback you had to do in production?”
“What is one code review rule that you would never remove from your team?”
“What is the last ‘small’ bug that cost your team a full day?”
One specific, low‑friction question beats ten generic ones.
Then:
Reply to every meaningful comment with something that adds value (your take, a question back, an extra detail)
Ignore “nice post” type comments or answer them only if you have something real to add
You are not farming comments.
You are starting conversations in public.
4. Go easy on links and growth hacks
For the next 90 days:
Share very few posts whose main point is “click this link”
If you must share one, make the post valuable by itself (your clear summary, your opinion, what you disagree with), and accept that reach may be lower — you are trading reach for clicks, by choice
Drop all automation, pods, and tricks. They are technical debt on your distribution.
Think like an engineer: remove flaky dependencies before you debug the core system.
5. Commit to a boring schedule
You don’t need to post every day. In fact, several data sets suggest that over‑posting correlates with worse performance per post.
A sane plan for a busy developer:
2 posts per week (for example, Tuesday and Thursday)
1 “big” piece per week (carousel or long text)
15–20 minutes per weekday commenting thoughtfully on other people’s posts in your niche
Do this for twelve weeks.
Then look at your saves, comments, and average dwell‑time proxy (long posts vs short):
If saves go up and comments get longer, you are training the right signals
Reach will follow with a delay, not instantly
No hacks. No pods. No magic template. Just 90 days of specific, on‑topic posts that real people actually read and save.
It’s not exciting.
But neither is refactoring a 2,000‑line legacy function — and you still do it when it matters.