687,000 LinkedIn posts · June 2025 to September 2026
LinkedIn algorithm changes in 2026: comment requests and thin posts lost reach
LinkedIn reach dipped a little for all posts in spring 2026. On July 27, it fell sharply for some kinds of posts only: posts that ask for comments, and posts with no concrete detail or lived experience.
Naïlé Titah8 min read
The essentials
Key takeaways
On July 27, 2026, all posts of the same 1,983 creators were only 2% lower (between 4% below and 1% above).
Comment requests dropped that week. “Comment KEYWORD” went from 10% above its author's average post to 47% below. Other comment requests went from 2% to 19% below.
Posts without substance (no concrete detail, no lived experience) fell further, from 13% to 23% below the average post. Sounding like AI already cost about 9% before July 27, with no extra cost since.
Reach takes days. A typical post has 8% of its final views after one hour, 23% after four hours and 72% after three days.
Posting hour, pods, fast replies and hashtags show no measurable effect on final reach. The first post after a break of two weeks or more gets 19% more impressions.
How to read the numbersEvery + or − percentage compares a post with the same creator's average post within 30 days before or after it: “−47%” means 47% fewer impressions than that creator usually gets. Two exceptions: formats are compared with photo posts, and link posts with posts without links.
Chapter 1The big picture
Did the LinkedIn algorithm change in 2026?
Yes, in several steps, and mostly for certain kinds of posts. On July 27, 2026, all posts of the same 1,983 creators were only 2% lower than before, somewhere between 4% below and 1% above their earlier level.
The one move measured on all posts together came in spring: by mid-May, posts were about 8% lower than the same creators' earlier posts. Our data starts in June 2025, so we can't rule out a seasonal dip.
Every other change we can date touched one kind of post at a time. LinkedIn announced feed updates in 2026, and some of our breaks sit close to their dates. The data can't show that one caused the other. How the feed ranks posts is covered in our guide to the LinkedIn algorithm.
Table 1 · LinkedIn algorithm changes in 2026, dated7 dates
| When | Source | What changed |
|---|---|---|
| Mar 12 | Announces a feed rebuilt around large language models, meant to “filter out engagement bait”. | |
| Mid-March | Our data | Photo posts gained ground: carousels' lead over photos fell from +21% to +9%. Text posts lost ground to photos too. |
| Spring | Our data | All posts, same creators: about −8% by mid-May. No 2025 data to compare. |
| Mid-May | Our data | Polls drop below the creator's photo posts: +9% before mid-May, −22% since. |
| Mid-May | Our data | Link cards cost less: −39% before, −19% since, against the creator's posts without links. |
| Jul 27 | Our data | Comment requests drop, and so do posts without substance. All posts together: −2%. |
| Jul 30 | Announces classifiers for “AI slop” on suggested content, and a button to report it. |
Chapter 2July 27, 2026
What changed in the LinkedIn algorithm on July 27, 2026?
Comment requests lost reach that week. From January 1 to July 26, 2026, a “Comment KEYWORD” post got 10% more impressions than its author's average post. Since July 27, it gets 47% fewer. Other comment requests (in English or French posts: “in the comments”, “👇”, “let me know”) went from 2% below the average post to 19% below.
LinkedIn said on March 12 that it would filter out engagement bait. Nothing moved in March: comment requests held their level until late July. Among 27 dates tested in 2025 and 2026, July 27 is the sharpest break for both kinds of request.
Table 2 · The last line of a post, before and since July 27, 20263 calls to action
| Last line | Jan 1 to Jul 26 | Since Jul 27 | Same creators, change |
|---|---|---|---|
| “Comment KEYWORD” | +10% | −47% | −45% |
| Other comment request | −2% | −19% | −29% |
| Closing question | −2% | 0% | −4% |
The gap opened within 15 to 30 minutes of publishing, on a small sample of posts published through MagicPost that we follow hour by hour. A closing question did not move. Formatting barely moved: none of the 17 formatting features we tested at that date, from bullet points to bold text, dropped more than 6%.
The keyword format has its own page: our study of LinkedIn comment bait.
Chapter 3Substance and AI
What changed on July 27 for posts without substance?
Among posts rated for substance, the ones that lost the most since July 27 have nothing concrete in them. A language model read 40,937 posts from 660 creators and noted whether each one held a concrete detail or a lived experience.
Posts without substance, with neither, were 13% below their author's average post from April to July 26, 2026, and 23% below since July 27. Among the same creators, they fell 15%.
- No concrete detail, no lived experience−13%−23%
- Both a concrete detail and a lived experience+10%+15%
- Random posts (control for the rows above)0%−1%
“AI and slop are not the same thing”
Hari Srinivasan, Chief Product Officer at LinkedIn, in his post announcing AI slop classifiers on July 30, 2026.See the post
Our data is consistent with that distinction. With the same substance and equally generic writing, a post that reads more like AI already got about 9% fewer impressions before July 27, between 6% and 12% fewer (30 points more on the model's 100-point AI score; this score was added to the analysis after the fact).
Since then, we can't detect any extra cost. The AI score and the generic score move closely together, so how the cost splits between them stays unclear.
On August 19, LinkedIn said members now see less of what it classifies as AI slop. That is LinkedIn's own measure, on its own definition, and it can't be compared with ours.
Chapter 4The first hour
Is a LinkedIn post decided in the first hour?
No. On 28,852 posts measured hour by hour, a typical post had 8% of its final views after one hour, 23% after four hours and 72% after three days. A post keeps growing slowly for weeks after that.
The posting hour changes the start. At one hour, depending on the time slot, a post has from 21% fewer to 11% more views than usual for its author. By the last reading, the hour makes no difference beyond 2% either way.
Chapter 5Rules of thumb
Which LinkedIn algorithm tips still work in 2026?
Few of them move final reach. Early comments from a pod make a post start faster, with 31% more views at one hour, and it finishes no higher. Replying to comments within the hour, commenting on other posts before publishing and adding one to three hashtags show no measurable effect.
The first post after a break of two weeks or more gets 19% more impressions than the creator's average post. The weeks that follow run a little lower, and we can't tell whether the break is the reason.
What to do with your next posts
Three moves, each backed by a number from this study.
- −47%since Jul 27
A “Comment KEYWORD” line now costs about half your usual impressions, and a plain comment request 19%. A real closing question carries no such cost: it has stayed level with the average post since July 27.
- −23%without either
Posts with both now run 15% above their author's average post. Posts without substance sit 23% below, down from 13% below before July 27.
- +19%after a 2-week break
A typical post gets 72% of its views in three days, and the hour you post changes the finish by 2% at most. The first post after two weeks off gets 19% more impressions.
How often to post on LinkedIn →
About this study
The data
687,000 posts
From 16,000 creators, published June 2025 to September 2026, brought together from our earlier studies. Hour-by-hour readings come from 28,852 posts by 1,294 creators who publish through MagicPost, the only posts we can follow hour by hour.
The method
Each post is compared with the same creator's average post: the mean, on a log scale, of their ordinary posts within 30 days before or after it. A random sample of posts lands at about 0% against this baseline. To date a change, we compare the same creators before and after a date, then repeat the test at many other dates in 2025 and 2026. Ranges come from resampling creators. A language model rated posts for detail, lived experience, generic writing and AI likeness; a second model rerated part of them.
The limits
- We don't see how LinkedIn ranks posts, so we date changes and measure their size without claiming a cause. LinkedIn's announcements are given for their dates only.
- We measure impressions, not saves or reading time.
- The spring dip has no seasonal check: our data starts in June 2025.
- September 2026 posts were still growing when measured. We will check the July 27 changes again in December 2026.
- Hour-by-hour readings exist only for posts published through MagicPost. This study names no creator and never shows a creator's impressions.
FAQ
Did the LinkedIn algorithm change in 2026?
Yes. MagicPost data on 687,000 LinkedIn posts dates the sharpest break to July 27, 2026: comment requests, and posts with no concrete detail or lived experience, have gotten fewer impressions since. All posts together barely moved that week (−2%, same 1,983 creators). All posts dipped about 8% in spring, a larger drop for everyone than on July 27.
Did LinkedIn reach drop in 2026?
A little, for all posts: about 8% lower by mid-May 2026, and only 2% lower on July 27 (same 1,983 creators), in MagicPost data on 687,000 LinkedIn posts. The big drops hit specific posts: since July 27, “Comment KEYWORD” posts get 47% fewer impressions than their author's average post.
What happened to the LinkedIn algorithm on July 27, 2026?
Comment requests lost reach on July 27, 2026. “Comment KEYWORD” posts went from 10% above their author's average post to 47% below, and other comment requests from 2% to 19% below, in MagicPost data on 687,000 LinkedIn posts. Closing questions did not move, and formatting barely did.
Do AI-sounding posts lose reach on LinkedIn?
Sounding like AI already cost about 9% of impressions before July 27, 2026, at equal substance, and MagicPost detects no extra cost since. Posts with no concrete detail and no lived experience lost more: 23% below their author's average post since July 27, from 13% below. MagicPost doesn't see how LinkedIn ranks posts, so it can't say why.
Is LinkedIn reach decided in the first hour?
No. A typical LinkedIn post has 8% of its final views after one hour, 23% after four hours and 72% after three days, measured hour by hour on 28,852 posts published through MagicPost. The posting hour changes the start, and the final reach by 2% at most.
Do hashtags, pods or replying fast still help on LinkedIn?
None of them shows a measurable effect on final reach. Posts with one to three hashtags sit 5% below the average post since July 27, 2026, within the margin of error (−14% to +5%). Early comments from a pod speed up the first hour, and the post finishes no higher. Replying within the hour shows no measurable gain.
Do links still hurt reach on LinkedIn in 2026?
Yes, though link cards to outside sites cost less reach than before. Since mid-May 2026, a link card sits 19% below the author's average post without links, against 39% below before. A URL typed in the text went from 12% to 21% below, and a “link in the comments” line from 8% to 16% below.
Do polls still get reach on LinkedIn in 2026?
Less than before. Since mid-May 2026, a poll gets 22% fewer impressions than the same creator's average photo post, against 9% more earlier in 2026, in MagicPost data on 687,000 LinkedIn posts. Carousels kept a smaller lead over photo posts: it fell from 21% to 9% in mid-March.
Cite this study
Charts and numbers are free to reuse with a link to this page.
MagicPost (2026). “LinkedIn Algorithm Changes in 2026: What the Data Shows.” Study of 687,000 LinkedIn posts. https://magicpost.in/studies/linkedin-algorithm-changes