How Does the LinkedIn Algorithm Work in 2026? 116,000+ Posts Analysed

You already know the rules the LinkedIn algorithm supposedly rewards. Post Tuesday morning. Use three to five hashtags. Keep it short. Never put a link in the post. End with a question. Open with something contrarian. And lately, a new one: LinkedIn changed the algorithm this year and nothing works any more.

Almost none of it is backed by evidence. And this is not stale folklore from 2019: Hootsuite, Sprout Social, Buffer and the rest republished these rules within the last year. Under each tactic below we link to who is making the claim right now, so you can see exactly what is being recommended before we test it.

So we tested it, first in July 2026 and again in October, against 116,000+ posts from 2,400+ B2B creators. We can't see inside the algorithm. Nobody outside LinkedIn can. But we can measure what it actually rewarded, and whether that changed. Here is everything we found.

Change in reactions, compared with the same creator's other posts

Worth doing

• Tell a personal story+33%

  Worked for 3 in 4 of the creators we tested

• Attach an image+35%

  Worked for 73% of creators

• Use "I", "my" or "we" in the first line+19%

  Worked for 69% of creators, in every format

Don't bother

• One or two hashtagsNo change
• "Unpopular opinion…" openersNo change
• Leading with a statisticNo reliable change
• Hitting a target post lengthNo magic number
• Picking the right day to postNo change
• Three or more hashtags−7%
• Ending with a question−7%
• Writing at the reader ("you should…")−10%
• Opening with a question−17%
• "Comment KEYWORD to get it"−24% since July

How we tested the LinkedIn algorithm

One rule separates this from most LinkedIn "studies", and it is worth thirty seconds.

Say you want to know whether hashtags work. The obvious test is to compare posts that use them against posts that don't.

The problem: the people who use hashtags are not the same people who don't. Maybe they are newer. Maybe they post more. Maybe they have smaller audiences. Whatever you find, you have measured the people, not the hashtags.

So we did it the other way round. We found creators who sometimes use hashtags and sometimes don't, then compared their hashtag posts against their own hashtag-free posts. Same person, same audience, same writing ability. The only thing that changes is the hashtag.

A tactic only counts as working if the same person does better with it than without it.

We ran the whole study twice. The first run, in July 2026, covered posts from July 2025 to July 2026. In October we collected every creator's newer posts and re-ran every test on the latest 12 months, October 2025 to September 2026. The numbers on this page are the October ones. Full detail on the sample is in the methodology at the end.

Did the LinkedIn algorithm change in 2026?

Not in the way people describe it. Across the same 2,100+ creators, the typical post earned about 10% fewer reactions from June to September than it had in spring. That is a slow slide, not a crash: there was no single week where engagement fell off a cliff, and it looked the same for small and large accounts in every region.

June to early July

90

Mid July to early August

90

August

86

September

83

Table view

A creator's typical post, compared with their own March to May 2026 posts

Category	Reactions, % of each creator's own spring level
June to early July	90
Mid July to early August	90
August	86
September	83

For scale, the same creators slipped about 7% over the same months of 2025, so roughly 7 points of this year's slide are new. Real, but nothing like the "reach has collapsed" posts suggest.

Two caveats. First, we measure reactions and comments, not views. Richard van der Blom's 2026 algorithm report, discussed on Creator Science in June 2026, puts reach for active creators down about 60% over two years while engagement fell only 20 to 25%. Fewer people may be seeing posts, while those who do still react. Second, some individual accounts genuinely did drop hard, about 1 in 12 of them. But a similar share drops hard every year, and the ones who fell furthest had mostly cut back on images and personal stories.

One thing really did change: engagement bait stopped working. MagicPost's October 2026 analysis reported that posts asking readers to "comment KEYWORD to get the guide" went from beating their author's usual views to roughly half of them after late July. We checked it on our own data. Before July those posts earned about 13% more reactions than the same person's other posts. Since July they earn about 24% fewer, and only 36% of the people still using them came out ahead.

We also checked whether accounts whose posts read as AI-written were singled out. They slid a little further than everyone else, but not by enough for us to be sure it's real. So if your numbers dipped this year, look at what you're posting before you blame the algorithm. The rest of this page says what to look at.

What actually works on LinkedIn in 2026?

Reactions, compared with the same creator's other posts

Attach an image

+35%

Tell a personal story

+33%

Use I, my or we in the first line

+19%

Use video

+15%

Include a concrete detail

+15%

Use a carousel

+11%

No change

Table view

Effect on a creator's own typical reactions

Tactic	Reactions vs their own posts	Change in reactions	Creators tested	Who benefited
Attach an image	1.35×	+35%	1,047	73%
Tell a personal story	1.33×	+33%	1,549	75%
Use I, my or we in the first line	1.19×	+19%	1,734	69%
Use video	1.15×	+15%	618	60%
Include a concrete detail	1.15×	+15%	1,000	62%
Use a carousel	1.11×	+11%	224	55%

Do personal stories work on LinkedIn?

Yes, and by more than almost anything else we've measured. Posts where the writer described something they actually did, saw or went through earned about a third more reactions than the same person's other posts. It worked for 3 in 4 of the 1,549 creators we could test.

This is new in the October run. To spot a story, we had an AI read every post from April to September 2026 (83,185 of them) and answer one question: is the author describing their own experience, or only giving advice? We then compared each person's story posts against their own advice posts.

Two related checks point the same way. Posts with at least one concrete detail, a real number, a named company or tool, a specific example, earned about 15% more. And posts with neither a story nor a detail, the generic advice that fills most feeds, earned about 13% less.

So the practical move: before you post, ask whether you could have written it without having been there. If anyone could have written it, add the part only you know.

Do LinkedIn posts with images get more engagement?

The same person's image posts earned 35% more reactions than their own text posts. It worked for 73% of the 1,047 creators we could test this way, in every region and at every audience size. In July the figure was 33%, so it has held steady.

That makes it the most reliable thing we found, and the least clever. No formula, no hook theory.

What kind of image? We don't know, and we won't pretend otherwise. Our data tells us whether a post had an image, not what was in it. That's the next study.

What we can say is that the lift comes from having one at all. So the practical move is simply to stop publishing naked text. A screenshot, a photo of the thing you're describing, or a plain chart all clear that bar. What does matter is the shape you export it at: the LinkedIn image sizes that survive the mobile crop decide how much feed space you take.

Video (+15%) and carousels (+11%) helped, but less, and only for a little over half the creators who used them. Treat those as worth trying rather than as safe bets.

Two things actively hurt. Posts that were mainly a link share earned about 11% less than the same creator's text posts. And polls were the worst thing in the entire study, at 58% down.

Only 47 creators qualified for the poll test, so treat that as a strong hint rather than a settled number. But nothing in our data recommends a poll.

Polls are also the sharpest disagreement between our method and the standard one. AuthoredUp's algorithm report (updated June 2026) gives polls a 1.64x reach multiplier, the best of any format, and SalesRobot (March 2026) says they "double the engagement". Both compare across different people, and they measure reach rather than reactions. Compared within the same creator, on reactions, polls came dead last. We cannot both be describing the same tactic; the difference is that cross-creator numbers measure who uses polls, while ours measure what a poll does to your own numbers.

Do first-person LinkedIn posts perform better?

Posts whose first line used I, my, we or our earned about 19% more reactions than the same creator's other openers. That was tested on 1,734 creators, and 69% of them benefited, exactly as in July.

This one survived unusually hard checking. First-person posts weren't more likely to carry an image, and weren't any longer, than everything else. So it isn't some other effect in disguise. It held in every format: text, image and video alike.

It's the same lesson as the story finding, at the scale of one line. LinkedIn readers reward posts that sound like a person describing something that happened to them. Not a brand. Not a coach. A person.

Same subject, same information. The second version of each simply reports something the writer did. Our breakdown of LinkedIn hooks takes all five first-line archetypes one at a time, including the two that cost you reactions.

This is why Flypost drafts from your own material rather than generating filler. The thing that works is the thing only you could have written.

What doesn't work on LinkedIn? Ten tactics that failed

Every number below comes from the same test: each creator measured against their own posts.

Reactions, compared with the same creator's other posts

1–2 hashtags

No change

"Unpopular opinion…" hook

−2%

6+ hashtags

−6%

End with a question

−7%

3–5 hashtags

−8%

Write at the reader ("you")

−10%

Add a link

−13%

Open with a question

−17%

"Comment KEYWORD" (since July)

−24%

No change

Table view

The tactics that failed. Effect on a creator's own typical reactions.

Tactic	Reactions vs their own posts	Change in reactions	Creators tested	Who benefited
1–2 hashtags	1.00×	No change	907	51%
"Unpopular opinion…" hook	0.98×	−2%	398	47%
6+ hashtags	0.94×	−6%	699	45%
End with a question	0.93×	−7%	924	41%
3–5 hashtags	0.92×	−8%	971	44%
Write at the reader ("you")	0.90×	−10%	1,382	35%
Add a link	0.87×	−13%	1,366	40%
Open with a question	0.83×	−17%	685	31%
"Comment KEYWORD" (since July)	0.76×	−24%	120	36%

Do hashtags work on LinkedIn?

One or two hashtags did exactly nothing. Three to five cost about 8%. Six or more cost about 6%.

The advice, as currently published: Sprout Social's hashtag guide, updated May 2026, says "limit yourself to three to five hashtags per post", and Hootsuite recommends the same range. Neither cites data for it. Three to five is the exact range that cost the most in ours.

About a third of posts still use them. That is an enormous amount of effort going into something that, at its very best, does nothing at all.

This is also a good example of why the fair test matters. Compared across different people, one or two hashtags looked helpful. Compared within the same person, the effect vanished entirely.

Does LinkedIn penalise posts with links?

Half true, and wildly oversold. Hootsuite's algorithm guide, updated July 2026, says "the algorithm doesn't prioritize posts with outbound links", and a July 2026 Forbes piece puts the penalty at 60%.

What we measured: adding a link to an otherwise normal post cost about 13% of reactions, and 40% of creators actually did better with links. That's almost exactly the July figure, so the link penalty hasn't grown into the catastrophe people describe.

Worth separating two different things here, because they get muddled:

• A post that is mainly a link share, a headline and a URL, does clearly worse: about 11% down on the same person's text posts.
• A normal post that happens to contain a link costs about 13%, and often earns it back.

So it is a trade-off, not a rule. If the click is the point of the post, 13% is a fair price.

Does "comment KEYWORD to get it" still work on LinkedIn?

Not any more. "Comment GUIDE below and I'll send it to you" was a growth trick that genuinely worked: before July 2026, those posts earned about 13% more reactions than the same person's other posts. Since July they earn about 24% fewer.

It's the one tactic in this study that changed direction during 2026. Fewer than 1 in 100 posts use it, so most people won't notice. If you do, offer the resource with a plain link or a direct message instead.

Does ending a post with a question get more comments?

This is the most confidently repeated tip on LinkedIn. ConnectSafely's 2026 guide says "every post should end with an invitation to engage", and Hootsuite recommends asking your audience a question. Neither offers data. It is also the tip we tested hardest.

The typical post that ends with a question gains about half a comment.

Across 924 creators, the typical change was about half an extra comment and about 7% fewer reactions.

To be fair to the tactic: those posts were more likely to gain a comment than lose one, and the penalty has eased a little since July. But comments come in ones and twos, so you are still giving up something real for very little.

Should you open a LinkedIn post with a question?

17% down, the worst opener we tested. Only 31% of creators benefited. Questions underperform at both ends of a post, though the opening question costs far more.

Hook guides still teach the question opener in 2026, but on this one the data keeps piling up on the same side. AuthoredUp ranked question hooks last of the five styles it tested across 309,614 posts, and MagicPost measured a 34% penalty across 1.2 million. Three studies, three different methods, same answer.

Do "unpopular opinion" hooks work on LinkedIn?

No difference. 398 creators, 47% benefiting: a coin flip.

Postiz claims the "Unpopular opinion:" opener "can triple engagement". Across the 398 creators who actually use it, the coin flip is what we found instead.

Does opening with a statistic get more engagement?

A little, if the number is in your first line. Within the same person, a first line containing a number earned about 8% more reactions, across 1,772 creators. Making the number the very first word did nothing reliable (288 creators).

MagicPost's analysis of 1.2 million posts found first lines containing a number earn more likes (35 median against 26). We agree on the direction but not the size: theirs compares across different posters, so part of their gap measures who writes number-led hooks. Our July version of this page only tested the number as the opening word and called it a dud; testing a number anywhere in the first line is the fairer match for their claim. So get a specific figure into your first line if you have one, but don't contort the sentence to lead with it.

Should you write "you" in a LinkedIn post?

Openers built on you and your, such as "You need to stop doing this", cost about 10%. Only 35% of creators benefited.

What is the ideal LinkedIn post length?

There's no magic number, and the famous 1,300-character figure appears nowhere in our data.

The sweet-spot claims keep coming, they just cannot settle on a number. AuthoredUp (updated July 2026) says posts of 1,301 to 2,500 characters earn "27% higher engagement". Other current guides say 1,200 to 1,800, or 1,300 to 1,900. When every study lands on a different ideal length, suspect the method rather than your posts.

What we found: the same person's long posts (1,500 characters or more) earned about 14% more than their own short ones (under 500), in the same format. A small bump, and a fraction of the gap the raw numbers first suggested, because most of that turned out to be long posts carrying images. For text-only posts the picture is less settled: a coin flip in July, a small bump in October, and too few people (153) to be sure either way.

If you want an anchor rather than a rule: a typical post in our sample runs to about 1,000 characters. Our full post-length analysis walks through how this one came apart.

What is the best time to post on LinkedIn?

Start with what the advice industry itself publishes, because the disagreement is the story. Sprout Social (updated March 2026) says Tuesday, late morning to afternoon. Hootsuite (November 2025) says 4am to 6am on Tuesdays and Wednesdays. Buffer (July 2026) says Wednesday at 4pm. Each cites its own dataset of a million posts or more. Three enormous datasets, three different best times: that is what noise looks like when it is sold as signal.

The one thing they agree on is to avoid weekends, and that is the one claim our data could fair-test. For the same person, posting at the weekend instead of midweek changed engagement by about 3%. That is noise. Of the 1,211 creators who post on both, 54% did better at the weekend: close to a coin flip.

What is solid is where everyone crowds. 60% of US weekday posts go out between 8am and 2pm. Weekends are only about 1 in 10 of US and UK posts.

The tempting conclusion is to post into the empty room at the weekend. Our data says that doesn't work either. The gap closes once you compare a creator against themselves. Post when it suits you.

Our full guide to the best time to post on LinkedIn goes through the day-by-day numbers, why the published timetables all disagree with each other, and what to do with the effort you were about to spend on scheduling.

How do you get more engagement on LinkedIn?

If you change one thing, make it the first item.

1. Tell a story only you could tell. Something you did, saw or got wrong, with a real detail in it. The strongest effect we've measured.
2. Attach an image. The most reliable single lift in the study. Export it at the right LinkedIn image size so it isn't cropped.
3. Write the first line about yourself. I, my or we, not "you should".
4. Drop the hashtags. One or two are harmless but pointless. Three or more cost you.
5. Skip the poll and the "comment KEYWORD" bait. The worst format we measured, and the one tactic LinkedIn turned against in 2026.
6. Put the link in if the click is the point. It costs about 13%. Often that's worth it.
7. Stop engineering questions. Not at the start, not at the end. Ask when you actually want an answer.
8. Write to the length the idea needs. Around 1,000 characters is typical, but there is no target to hit.
9. Post when it suits you. Then spend the time you saved on the post itself.
10. Aim for three to four posts a week if total reach is what you're after. Expect each individual post to earn less.
11. Judge yourself against people your own size. A typical 10,000-follower account gets about 31 reactions a post, not the 139 a large account gets.

If you're choosing a tool to help with any of this, we compared 11 Taplio alternatives on their real prices, our own included.

How often should you post on LinkedIn?

Post more often. Each post will get fewer reactions, but your total for the week goes up sharply, and the weekly total is the number that grows an audience.

Reactions per postReactions per week

Under 1.5/week

31

12

1.5–4 per week

16

39

4+ per week

8

42

Each post earns about a quarter as much. The week earns over three times more.

Table view

Posting frequency for creators with similar-sized audiences. Accounts under 10,000 followers.

Group	Reactions per post	Reactions per week
Under 1.5/week	31	12
1.5–4 per week	16	39
4+ per week	8	42

Going from about once a week to four or more times a week cuts each post from 31 reactions to about 8. It lifts the week from 12 to 42. You lose per post and win overall, by more than three times. The same shape holds at every audience size.

So the trap is judging yourself on the wrong number. Per-post engagement drops when you post more, and it feels like something has gone wrong. Nothing has.

Can you post twice in one day on LinkedIn?

Yes, though the two posts do eat into each other a little. This one was fair-tested, by comparing each creator's posts that followed a short gap against their own posts that followed a long gap.

It runs in both directions. A post that follows another within a day earns about 10% fewer reactions. And the earlier post loses about 9% when another one follows within two days.

Should you vary your LinkedIn post formats?

The advice: mix up your formats to keep the algorithm interested. Hootsuite recommends posting "a combination of" content types, and AuthoredUp says mixing content types keeps followers engaged. Neither offers data for the switching itself.

No support at all. We compared each creator's posts that changed format from their previous post against those that repeated it. Switching was about 5% worse, across 2,000 creators tested, with only 44% benefiting.

Don't force variety for its own sake. Find what works for you, usually images, and keep doing it.

How many reactions is normal on LinkedIn?

The most common question we get isn't about tactics at all. It's whether your numbers are any good.

The typical B2B creator's post earns about 36 reactions and 3 comments.

5k–10k

26

10k–25k

31

25k–50k

44.5

50k–100k

73.5

100k–500k

139

middle half of creators median

The bands overlap: a top-quarter creator with 10,000 followers (56 reactions) beats a bottom-quarter creator with 50,000 (36).

Table view

Reactions per post by follower count

Followers	Lower quarter	Median	Upper quarter	Typical comments
5k–10k	12	26	49	2
10k–25k	15	31	56	3
25k–50k	21	44.5	80.5	5
50k–100k	36	73.5	162.5	10
100k–500k	58	139	270	16

Three things worth noticing.

Comments are the scarce currency. The typical creator under 25,000 followers gets between one and three comments per post. If one of your posts gets ten, that is genuinely good, and it puts all that "end with a question" advice in perspective.

Your reach per follower falls as you grow. Reactions and comments together come to roughly 0.4% of followers for accounts under 10,000, dropping to about 0.1% above 100,000. A small account works its audience around four times harder than a large one. Those figures are approximate because follower counts were captured early in 2026 and have grown since.

The spread inside every band is enormous. That is what the shaded bars are really showing. The gap between a weak and a strong creator at the same follower count is routinely five to ten times. Your audience size sets a floor, not an outcome.

Methodology

For the sceptics, and for anyone citing this.

Who we looked at. 2,500 English-speaking B2B creators, spread across region (US, UK/CA/AU, EU, rest of world) and follower band (1,000–500,000), with country caps so no single market could dominate. Individuals only, no company pages.

What we collected. Two sweeps of the same creators. The first, in July 2026, captured about almost 130,000 posts covering July 2025 to July 2026, of which 86,000+ went into the original analysis. In October 2026 we collected the newer posts of the 2,100+ creators still posting, bringing the total to 180,000+ posts. The numbers on this page cover the latest 12 months, October 2025 to September 2026: 116,000+ posts from 2,400+ creators. Reposts are excluded throughout, because the engagement belongs to whoever wrote the original.

Checking the re-run. Before using the October numbers we rebuilt every test and ran it on the July data first. Each one landed within a few percentage points of what we had published, so the October figures use exactly the same method.

What we measured. Reactions and comments only. We never measured impressions. Nobody outside LinkedIn can, and any study claiming otherwise is estimating. Nothing counts until a post is at least 14 days old, because the numbers are still moving before that. We use typical figures (medians) rather than averages throughout, so one viral post can't drag a result upward.

How we tested. Every cause-and-effect claim compares a creator against their own posts, with at least three posts on each side (four for first lines). Creators who tied exactly are excluded from the "who benefited" percentages rather than counted as losses. Sample sizes are stated with each finding and vary because not every creator qualifies for every test.

Stories and details. Whether a post tells a personal story or contains a concrete detail was judged by an AI model reading each post from April to September 2026 (83,185 posts), so those three findings cover six months rather than twelve. The "comment KEYWORD" result compares April to early July against mid-July to September, because it changed direction in between.

Timing. Post times come from LinkedIn's internal post IDs rather than the "3 months ago" labels, which are unreliable, then converted to the local time of each author's country. Author-local is a proxy for audience-local, which is a real limitation on any timing claim.

What we're less sure about. Follower counts date from early 2026, so the reach-per-follower figures are mild over-estimates. The poll and carousel findings rest on small samples (47 and 224 creators), and so does "comment KEYWORD" (120 creators since July). Posting frequency is the one finding we couldn't fair-test. The double-posting result covers people who occasionally post twice a day. Habitual double-posters are excluded by the method, and three or more posts a day is untested. We also can't see comment threads, so anything involving replies or first-comment links is outside what this data can answer. Several other tactics didn't produce a result we trusted, and we left them out.

Spot a flaw? We would genuinely like to know. Tell us and we will correct it here.