Is LinkedIn organic reach declining? Tracing the numbers to source

LinkedIn organic reach is declining, and the number everyone quotes for it is the one we could not verify. Four independent datasets with published methodologies all point the same way: Metricool measured impressions down 10% year over year across 673,658 posts, AuthoredUp reported average reach down 34% across 621,833 posts, Socialinsider measured video views down 36%in every follower tier across 1.3 million posts, and Buffer measured LinkedIn's median engagement rate down about 5%. The widely repeated 50% figure comes from a report that sells for 1,500 euros and up, publishes no sample size on its own website, and has never made the underlying data public. Direction: confirmed. Magnitude: contested, and the most dramatic version is the least checkable.

Is LinkedIn organic reach actually declining?

Yes. This is one of the rare cases where independent panels using different metrics on different populations all move in the same direction, which is about as much agreement as social media measurement ever produces.

Read the metric column carefully, because it does most of the work. Impressions, reach, video views and engagement rate are four different things. A 10% impressions drop and a 34% reach drop can both be true at once on different panels. They are not two estimates of one number.

Where does the “LinkedIn reach is down 50%” figure come from?

It comes from the Algorithm Insights report published by Richard van der Blom and Just Connecting. It is the most cited source in this entire subject area, and it is a commercial product.

Here is precisely what we could and could not verify.

We want to be careful here, because there is a lazy version of this argument that we are not making. We are not saying the research is wrong. We have not seen it. Practitioners we respect cite it, including AuthoredUp, which references it for its first-hour engagement claim and for a feed composition breakdown. A paid research product is a legitimate thing to sell.

What we are saying is narrower and it matters for how you use the number. When a figure's chain of custody ends at a paywall, nobody outside the purchase can check the sample, the panel, the definition of a view, or the date window. Repeating it as established fact turns an unaudited claim into consensus. That is how “reach is down 50%” became something everybody knows and nobody has checked.

Why do the measured studies disagree on how much reach has fallen?

Four reasons, and none of them means any of the studies is lying.

• Different metrics. Impressions count displays. Reach counts distinct people. Video views count a different event again. LinkedIn itself defines impressions as “the number of times your post was shown” and members reached as “the number of distinct members and Pages that saw your post”, which excludes repeat displays. Those two numbers can move apart.
• Different panels. Metricool, Buffer and AuthoredUp all measure their own customers, who are people serious enough about posting to pay for a tool. Socialinsider measures company pages exclusively. None of them is a random sample of LinkedIn.
• Different windows. Metricool compares two two-month windows a year apart. Socialinsider compares full calendar years. AuthoredUp reports a point-in-time figure for 2025. A steep quarter inside a flat year looks very different depending on where you cut.
• Mean against median. Engagement and reach are heavily skewed. A mean falls faster than a median when the top of the distribution cools, and most studies do not say which they report.

This is the same denominator-and-population problem that makes every published LinkedIn engagement benchmark disagree with every other one. It is worth internalising once, because it explains most of the apparent chaos in this field.

Is engagement falling as fast as reach?

No, and the divergence is the most interesting finding in the whole set. Metricool's year-over-year comparison:

• Impressions per post: down 10%
• Overall engagement: up nearly 14%
• Likes: down 13%
• Comments: down 17%
• Shares: down 10%
• Clicks: up 5%

Every visible interaction type fell. Total engagement rose. That is only possible because the denominator, impressions, fell faster than the numerator, and because clicks, which are counted as engagement and are not publicly visible, went up.

Metricool calls this category invisible interactions and defines it as “actions like clicks, carousel swipes, video views and link taps” that “signal user intent even if not publicly visible”. If your like counts have fallen over the past year and you concluded your content got worse, this is the alternative explanation: fewer people saw it, and those who did behaved differently rather than less.

It also lines up with what LinkedIn itself has been building. Its published feed models predict long dwell and contributions as separate objectives, with click and skip modelled alongside them. A feed tuned on quiet attention will produce fewer public reactions per impression by design. We went through the published engineering work in what LinkedIn has published about dwell time.

Which formats lost the most reach?

AuthoredUp published median reach changes by format across its 621,833 post dataset, and the ranking is not the one most people expect.

Figures from AuthoredUp's September 2025 analysis. The multiplier column is that format's reach relative to a baseline post.

Two things worth noticing. Polls lost the least reach and carry the highest multiplier for personal profiles, which matches Metricool's finding that polls deliver the highest average impressions on company pages despite being under 1% of all posts. A format almost nobody uses is holding up better than the formats everybody uses. If you have written off LinkedIn polls, the current data does not support that.

Video is the other odd case. Socialinsider recorded video views falling 36% across every follower tier while reporting video engagement rate rising to 6.00%. Metricool found video underperforming carousels and images in 2026, reversing its own 2025 finding. Fewer views, higher rate per view. That is a decline in distribution rather than a decline in the format, and it is worth keeping in mind before you rebuild your video strategy around a single study.

Is follower growth slowing too?

Sharply, and most sharply for the largest accounts. Socialinsider's audience growth rate by follower tier, comparing 2024 with 2025:

The top tier lost roughly two thirds of its growth rate in a single year. Metricool measured the same thing from a different angle and found that only 7% of LinkedIn company pages grew enough to move into a higher follower tier.

For anyone building an audience, that reframes the target. If you are in the 1,000 to 5,000 band, a 24.50% annual growth rate is the current benchmark, not the 40.75% that a 2024 article would have told you. Our guide to the first 1,000 followers was written for the earlier environment and the arithmetic there is now conservative.

Are personal profiles losing reach as fast as company pages?

The measured evidence says the two are much closer than the folklore suggests, at least on impressions. Metricool covered both account types in the same window and recorded 817.67 impressions per post for personal profiles against 812.64 for company pages. Effectively identical.

The gap is in response, not distribution. Profiles ran a 2.60% engagement rate against 1.74% for pages, roughly 63% higher, on 14.44 interactions per post against 14.16. Metricool also found that pages collect more shares while profiles collect substantially more comments.

So the common advice to abandon the company page and post from a profile is not, on this evidence, a reach argument. It is a response argument, and the mechanism is that people answer people. If comments matter to you, and in our data they are the signal that separates the top decile from everything else, the profile is the better instrument. That does not make it a shield against a platform-wide decline in impressions, because both account types are experiencing it.

Does posting more often make up for the lost reach?

The data available runs against that instinct. Socialinsider recorded company pages increasing image posts from an average of five to seven per month and video from two to four per month between 2024 and 2025, across the same period in which average video views fell 36% in every follower tier. More output, less distribution each.

Metricool found the sharpest counter-example. Accounts above a million followers cut posting frequency by 50.42% and saw engagement rise 94.76%, while accounts between 2,000 and 10,000 followers grew engagement 23.17%. One study, one correlation, and a very large account is not a normal account. But it is the opposite of what a volume strategy predicts, and it is the only measurement of the question we could find.

The reading we would take from it is not “post less”. It is that cadence is not the lever for reach recovery, and that treating it as one produces a lot of mediocre posts that pull your own median down. Our take on the trade-off, with the arguments for higher cadence included, is in how often to post on LinkedIn.

Has LinkedIn said anything about organic reach?

Not about volumes. LinkedIn has published no statement about how much organic reach accounts are getting, and we would not expect one. What it has published is engineering detail, and that detail is informative if you read it as a description of incentives rather than an announcement.

LinkedIn's February 2026 feed ranking paper names two optimisation targets: long dwell, meaning a member stays on a post past a threshold that varies by post type, and contribution, meaning likes, comments or shares. Its March 2026 engineering post describes a retrieval stage built on LLM-generated embeddings feeding a generative recommender that reads “more than a thousand of your historical interactions”.

A feed that retrieves and ranks on predicted personal relevance, at that level of granularity, distributes less broadly by construction. Posts that the model does not expect a specific member to dwell on do not reach that member. That is not a penalty and it is not a conspiracy. It is what relevance ranking does. The full picture of how the ranking stack fits together is in our guide to the LinkedIn algorithm.

What can our own data say about this, and what can it not?

We should be straight about this: our dataset cannot measure a reach decline. It has no impressions field, no reach field, and its engagement counts are lifetime totals captured at a single scrape, so we cannot build a year-over-year series from it. Anybody who tells you their engagement-only dataset proves a reach trend has skipped a step.

What our 12,988 posts from 65 creators do show is the shape of the distribution that any decline is acting on, and that shape explains why a platform-wide drop feels so much worse than the percentages suggest.

• Median engagement rate: 0.40 per 1,000 followers. 75th percentile: 1.27. 90th percentile: 5.95.
• Top-decile median: 10.26. Bottom-half median: 0.15.
• Median post: 122 reactions and 12 comments. Top decile: 235 and 72. Bottom half: 80 and 5.

Engagement on LinkedIn is not normally distributed. Most posts sit in a narrow band near the floor and a small number break out by an order of magnitude. When platform-wide distribution tightens, the posts that vanish first are the ones already near the floor, which is most of them. The average creator does not experience a 10% impressions decline as a 10% haircut on every post. They experience it as their weakest posts going quiet, which reads as a cliff.

One more figure from our cohort makes the point sharper. Only 40 of the 65 creators produced even one top-decile post across the whole dataset. Breakouts were already concentrated in a minority of accounts before any of these declines were measured. A tightening feed does not redistribute that concentration evenly. It removes the marginal posts that were only ever reaching people through the loosest part of the distribution, which is the part that relevance ranking is designed to cut. If you want the distribution argument in full, it is in what the data says about going viral on LinkedIn.

How do you tell a platform-wide decline from your own problem?

A procedure you can run in half an hour.

1. Pull your last 30 posts and the 30 before those. Record impressions, members reached, reactions and comments for each.
2. Compare medians, not averages. One breakout post will drag an average around and tell you nothing.
3. Compute the percentage change in your median impressions. If it is somewhere in the 10% to 35% range, you are moving with the platform and there is no mystery to solve.
4. If it fell much further, check format mix first.A quiet drift toward shared links will do this on its own. Link posts sit at the bottom of every format table we have seen, at 3.25% in Socialinsider's data and 3.81% in Buffer's, and at the bottom of our own rate table too.
5. Check cadence. Socialinsider recorded pages raising image posts from five to seven per month and video from two to four while views fell. More posts do not recover lost distribution and can dilute it.
6. Check comments separately. Comments are the scarcest signal in our data and the one that separates tiers hardest. If impressions fell but comments per impression held, your content is fine and your distribution changed.

If the drop was sudden rather than gradual, the diagnostic order is different and we set it out in why LinkedIn reach suddenly drops. If a single post died on arrival while the rest are normal, start with why a LinkedIn post gets no views instead.

What still works when reach is falling?

Three things survive contact with all of the data above, and all three are unglamorous.

Formats that hold the reader on the post.Documents and carousels top every format table with a published methodology: 7.00% in Socialinsider's, a 21.77% median in Buffer's, and a 1.45x reach multiplier in AuthoredUp's. Polls lost the least reach of any format in AuthoredUp's data and carry the highest multiplier for personal profiles.

Posts that earn comments rather than reactions.In our cohort the median top-decile post earns 72 comments against 5 for the median bottom-half post, a far wider gap than the reaction gap of 235 against 80. Comments are also the metric falling fastest platform-wide in Metricool's data, down 17%, which makes them scarcer and therefore more discriminating. Our comments playbook covers how to write endings that produce them.

Consistency measured against yourself. Every published benchmark on this page will be wrong within a year. Your own trailing median will not be.

What are the limits of everything on this page?

• We cannot measure reach. No impressions field, no reach field, single scrape. Every trend figure here belongs to somebody else and is linked.
• Vendor panels are not random samples. They measure paying customers of social tools, who post more and more deliberately than the median member.
• Company pages and personal profiles are different populations and several of these studies cover only one of them.
• The 50% claim may well be accurate. We are not asserting it is false. We are reporting that we could not check it, and that nobody quoting it appears to have checked it either.
• Our distribution figures are correlational and come from 65 established creators with lifetime-cumulative engagement counts, describing the visible hook only.

Is LinkedIn organic reach declining? Restated

Yes, and the honest range is roughly 10% to 35% depending on which metric and which panel you trust. Impressions are down 10% in Metricool's 673,658 post study. Reach is down 34% in AuthoredUp's 621,833 post study. Video views are down 36% across every follower tier in Socialinsider's 1.3 million post study. Follower growth for the largest pages fell from 21.60% to 6.40% in a single year.

The 50% figure is a different kind of claim. It may be right. It is not checkable, because the document behind it is a paid product whose public pages carry no sample size and no findings. We think the useful move is to stop arguing about the headline number and start comparing your own last thirty posts against your own previous thirty, which is the method behind our study of 12,988 LinkedIn posts and the only benchmark that will still be true next quarter.