The LinkedIn algorithm decides what each member sees by first retrieving posts that match their profile and interests, then ranking them with a model that reads their recent history and predicts whether they will dwell on a post or react, comment or share. Low-quality, bait-driven and off-topic posts are held back. Follower count matters less than relevance.
That is the short version. The longer version changed a lot in 2026. In March, LinkedIn's engineering team described how it had replaced a patchwork of retrieval systems with a single language-model system and moved ranking to a sequence model that looks at more than a thousand of your past interactions. Its newsroom published the same week a list of what it is now pushing down: "comment to agree" prompts, recycled posts and engagement pods. This guide walks through what LinkedIn has actually said, what independent studies of millions of posts show, where the two disagree, and how to turn all of it into a weekly routine.
How the LinkedIn Algorithm Works: Key Findings
If you only read one section, read this one. Each point links to the source it comes from.
- The feed was rebuilt in 2026. LinkedIn's engineering post of 12 March 2026 says the ranker "processes more than a thousand of your historical interactions" and that new posts get their embeddings "in near-real time".
- Dwell time is a first-class signal. The production ranker described in LinkedIn's February 2026 Feed SR paper is trained on two main targets: Long Dwell and Contributions (a like, comment or share). In online tests it lifted time spent by 2.10% over the previous model.
- Bait is now named and demoted. LinkedIn's March 2026 newsroom update lists "comment to agree" prompts, videos that do not match their text, recycled content, engagement pods and comment automation as things it is reducing.
- Documents and multi-image posts lead on engagement rate. Socialinsider's analysis of 1.3 million business posts puts native documents at 7.00% and multi-image at 6.45% average engagement in 2025, against 3.25% for link posts.
- Visible engagement is falling, invisible engagement is rising. Metricool's 2026 LinkedIn study of 673,658 posts found likes down 13%, comments down 17% and shares down 10% year over year, while overall engagement rose nearly 14% thanks to clicks, swipes and video views.
- People beat Pages. The same Metricool study found personal profiles earn a 63% higher average engagement rate than Company Pages, and RevBoss's 2026 benchmark found founder posts reached a median 520 impressions against 244 for company pages.
- Hashtags correlate with lower reach. In Ordinal's study of 270,306 posts, posts with no hashtags averaged 6,632 impressions, while posts with 11 or more averaged 502.
What's Changed in the LinkedIn Algorithm in 2026
Most "algorithm update" articles recycle the same myths every year. This year there is real, documented change, and it is worth separating into what LinkedIn built, what LinkedIn said it would reward and punish, and what the data shows creators are experiencing.
A new retrieval system built on a language model
Before 2026, LinkedIn assembled each feed from many separate candidate sources. Its engineering team wrote that "each source had its own infrastructure, index structure, and optimization strategy". The replacement is one dual-encoder system that uses a language model to turn both members and posts into embeddings and then matches them by similarity. The research behind it, published in October 2025, fine-tuned Meta's LLaMA 3 as the encoder and reports retrieving 2,000 candidates from a pool of hundreds of millions of posts. In its online test, the biggest gains went to members with fewer connections, where daily professional interactions rose 1.17%.
For creators, the practical meaning is that LinkedIn now reads what a post is actually about. The engineering post says the model "better understands what a post is actually about" and uses world knowledge to connect related topics, giving the example of linking electrical engineering to small modular reactors. Keyword stuffing and hashtags matter less when the system can understand the topic from the text itself.
A ranker that reads your history as a sequence
Once candidates are retrieved, a Generative Recommender model ranks them. Instead of scoring each post in isolation, it reads a member's interactions in order. The Feed SR paper says the model keeps "the most recent T=1000 impressions" per member, replaced an older DCNv2-based ranker and is "currently the primary member experience on LinkedIn's Feed". Online, it improved time spent by 2.10% and daily active users by 2.38%.
A sequence model notices trajectories. If someone has spent the last two weeks reading about pricing strategy, posts about pricing strategy will rank higher for them now, even if they never followed you. That is why posting consistently about a defined set of topics compounds: you keep showing up in the sequences of people whose interests match yours.
An official list of what gets pushed down
The same week, LinkedIn's newsroom published How LinkedIn Is Improving the Feed. It names the patterns being reduced: engagement bait such as "comment to agree" prompts, videos that do not match the text, repetitive low-substance and recycled posts, click-driven posts, and engagement pods, comment automation and unauthorised third-party tools that fake conversations. It also says the feed now uses "signals you choose to share, like your industry, skills, experience, and geography" alongside engagement over time, and that LinkedIn is testing an Interest Picker for new members.
What the data shows creators are experiencing
Two independent datasets point the same way. Metricool's 2026 study (January and February 2025 compared with the same months of 2026) found visible reactions falling while clicks, carousel swipes and video views carried overall engagement up. Socialinsider's benchmarks show average video views for Pages with 10,000 to 50,000 followers dropping from 1,000 in 2025 to 585 in 2026, while video posting frequency doubled from two to four a month. More video is chasing the same attention, and generic video is losing.
| Change | When | What it means for your posts |
|---|---|---|
| Long Dwell classifier added to ranking | October 2024 | Passive reading counts, even without a reaction |
| LLM dual-encoder retrieval research published | October 2025 | Topic is understood from your words, not hashtags |
| Feed SR sequence ranker paper published | February 2026 | Members' recent interests shape what ranks today |
| Engineering rebuild and newsroom update | 12 March 2026 | Bait, pods and recycled posts are named for demotion |
| Metricool 2026 LinkedIn study | April 2026 | Likes and comments down, clicks and swipes up |
The Ranking Pipeline, Stage by Stage
LinkedIn has never published a single diagram of its whole feed, but its engineering posts, papers and help pages describe the pieces clearly enough to assemble one. Think of it as four stages that run every time a member opens the app.
Stage 1: Retrieval, from millions of posts to a shortlist
The retrieval system compares a member's embedding with post embeddings and pulls the closest matches. The engineering post puts retrieval at "sub-50ms" across an index of millions of posts. For new members, it says the language model "can deduce likely interests and relevant content areas based on member-provided information", which is why a complete profile and clear headline matter on both sides of the feed: they shape who you see, and who sees you.
Stage 2: Ranking, predicting dwell and contributions
The shortlist is scored by the sequence ranker. The Feed SR paper appends about 512 candidates to a member's history and scores them in one pass. Its main prediction targets are Long Dwell, meaning time on a post beyond a threshold that depends on the post type, and Contributions, meaning a like, comment or share. The older production model also predicted clicks and skips.
Stage 3: Quality and policy filters
LinkedIn's help page on LinkedIn relevance says its algorithms "filter out or taper distribution of low-quality and unsafe content". Its guidance on content recommendations to avoid lists five categories that may not be shown beyond your connections and followers: promotional or off-topic posts, engagement bait, unoriginal posts, unconstructive posts, and adult or sensitive content. This is a filter on reach, not on publishing. Your network can still see the post; the wider feed will not.
Stage 4: Feedback and expansion
Every reaction, comment, long read, skip and "I don't want to see this" flows back into the member's sequence. The feed preferences help page describes how members can mark a post as not interesting, unfollow or mute. Posts that earn dwell and contributions from the first people who see them keep getting retrieved for similar members, which is how a post reaches second- and third-degree audiences.
| Stage | What happens | What you influence | Source |
|---|---|---|---|
| Retrieval | Post embedding matched to member interests | Clear topic in plain words, consistent subject matter | LinkedIn Engineering, March 2026 |
| Ranking | Sequence model predicts Long Dwell and Contributions | A hook that earns the read, a body worth finishing | Feed SR paper, February 2026 |
| Filtering | Low-quality or bait posts kept to your own network | No bait, no promotion without insight, original takes | LinkedIn help and guidance |
| Feedback | Reactions, reads and hides update members' histories | Replying to comments, avoiding off-topic posts | LinkedIn Help, feed preferences |
Key takeaway: there is no single "LinkedIn score" for your post. There is a separate prediction for every member who might see it. Your job is to make the post obviously relevant to a specific kind of professional and worth their time once they start reading.
Ranking Signals by Type: Identity, Content and Activity
LinkedIn's own relevance help page groups the inputs into three types. This is the most useful official framing available, so it is worth working through each with what it means in practice.
Identity signals: who the viewer is, and who you are
Identity covers profile details such as location, workplace, skills and professional background. The 2026 newsroom post adds industry, skills, experience and geography as signals members "choose to share". This works in both directions. A viewer's profile decides which topics they are matched with, and your profile helps the system understand which topics you are credible on. A headline that says "Helping B2B SaaS teams fix churn" gives the model far more to work with than "Entrepreneur | Speaker | Dreamer".
Content signals: what the post is, and how it is doing
According to the same page, content signals include how often a post is viewed or engaged with, its topic, how recent it is, whether it comes from someone the viewer follows or is connected to, its language, and "how constructive or professional the conversation is". That last one matters. A comment section full of thoughtful replies is itself a signal, separate from the number of comments.
Activity signals: what the viewer does
Activity covers what the viewer reacts to, comments on and shares, who they interact with, how long they view content and which topics they have shown interest in. This is the part the sequence model is built around. It is also why relationship strength still matters: if someone regularly reads and replies to your posts, your next post has a strong prior in their feed.
Dwell time: the signal most creators underrate
LinkedIn has written about dwell time for six years. Its 2020 engineering post explains that dwell "on the feed" starts measuring "when at least half of a feed update is visible", and that it fed a P(skip) model which reduced the score of updates people were likely to scroll past. In October 2024, the team added a Long Dwell classifier, noting that "a significant portion of our weekly active users are passive consumers". Thresholds are calculated per content type, creator type and distribution method, because "a dwell time of 30 seconds could be considered 'long' for an image but not for a video".
The implication: a post that a few hundred people read carefully, with few reactions, can still be a strong post in LinkedIn's eyes. Measuring yourself only by likes misses most of what the ranker sees.
| Signal | Type | Officially documented? | How to improve it |
|---|---|---|---|
| Long dwell on the post | Activity | Yes, engineering blog and Feed SR paper | Open with a specific claim, keep paragraphs short, deliver on the promise |
| Likes, comments and shares | Activity | Yes, "Contributions" target | End with a question people can answer from experience |
| Skips | Activity | Yes, P(skip) model | Avoid vague first lines that look like every other post |
| Topic match | Content | Yes, help page and LLM retrieval | Name the topic in plain words in the first lines |
| Recency | Content | Yes, help page | Post when your audience is reading and you can reply |
| Conversation quality | Content | Yes, "constructive or professional" | Reply with substance, invite disagreement politely |
| Profile details | Identity | Yes, help page and newsroom | A headline and About section that state your field |
| Hides and "not interested" | Activity | Feed preferences help page | Stay on topic; do not post for everyone |
The LinkedIn Algorithm by Feed Surface
Not every LinkedIn view comes from the same place. The feed you see is a blend, and each part behaves differently.
The Top feed (the default)
LinkedIn's help page on feed sorting says Top posts are "personalized based on your activity and interests". This is where nearly all ranking applies, and it is where most of your impressions come from.
The Recent feed
Members can switch to Recent, which shows "updates in reverse chronological order from people you follow, without personalization". If someone sets Recent as their permanent view, "your interests and activity data will not be used to sort the feed or make recommendations". In practice few people do this, but it is the one place where posting time is the only factor.
Recommended posts from outside your network
The feed overview help page says the feed mixes posts from your network, companies you follow, sponsored content and "other posts that we believe you may be interested in", labelled with tags such as Trending or Recommended for you, and that this recommended content "is determined by LinkedIn's algorithms and editorial team". The Feed SR paper notes that an earlier language-model ranker did well on out-of-network recommendations but struggled with network posts, one reason LinkedIn combined approaches. Out-of-network reach is held to a higher bar: LinkedIn's guidance is explicitly about posts that may not be shared beyond your connections and followers.
Video in the full-screen player
Video has its own viewing experience on mobile. Digiday reported that LinkedIn told it total video viewership grew 36% year over year between 30 October 2024 and 29 January 2025, and that video creation was growing at "twice the rate of other post formats". Hootsuite's July 2026 guide says LinkedIn reported a third straight quarter of double-digit growth in video uploads in Q1 2026. More supply means more competition for each viewer, which matches the falling average views in Socialinsider's data.
How Each Format Performs in 2026
LinkedIn does not boost a format because it is a format. Formats change how people behave, and behaviour is what the ranker measures. Here is what the largest recent datasets show, with the caveat that each measures something slightly different.
| Format | Avg engagement rate 2025 (Socialinsider, Pages) | Median impressions 2026 (RevBoss, mostly founders) | Posts per month on Pages, 2025 |
|---|---|---|---|
| Native document | 7.00% | Not reported | 2 |
| Multi-image | 6.45% | 1,042 | 1 |
| Video | 6.00% | 565 | 4 |
| Single image | 5.30% | 529 | 7 |
| Text only | 4.50% | 508 | 1 |
| Poll | 4.20% | Not reported | 1 |
| Link | 3.25% | 542 (article links) | 4 |
Sources: Socialinsider, 1.3 million posts from 16,645 Pages, 2024 to 2025; RevBoss, 1,202 posts from 33 profiles, January to August 2026.
Text posts
Text-only posts sit mid-table on engagement rate but are the most common format for founders in RevBoss's sample, with 690 of 1,202 posts. Their reach depends almost entirely on the first two lines, because the "see more" click is the first dwell signal. Our LinkedIn post formula covers how to write those lines, and the LinkedIn character counter shows where the cut-off falls before you publish.
Single images
Images are the workhorse format for Pages, with seven a month on average in Socialinsider's data. LinkedIn's own Pages best practices say images "typically result in a 2x higher comment rate". The image should carry information, such as a chart, a screenshot or a before-and-after, rather than a stock photo. Our guide to LinkedIn image posts covers sizes and design, and the LinkedIn image resizer exports the right dimensions.
Multi-image and document posts
These are the two formats where readers swipe, and swipes are exactly the "invisible interactions" that Metricool found driving engagement up. In Socialinsider's Q2 2026 quarterly figures, multi-image reached 6.90% and native documents 6.60%. Both build long dwell naturally, because each swipe is another moment of attention. We compare them with text in LinkedIn carousel vs text post.
Video
Video is now the most-used format on personal profiles, according to Metricool, yet it underperforms carousels, image posts and multi-image content there. LinkedIn measures a video view as two or more seconds of continuous playback with the video at least half visible, according to Digiday's reporting, and Long Dwell thresholds for video are set relative to other videos. A talking-head clip that loses people in five seconds is competing against every other video in the same bucket. The newsroom's warning about "videos that don't match the text" also matters: a video pasted under an unrelated caption is a named demotion pattern.
Polls
Polls are the outlier. Metricool found they make up less than 1% of posts but deliver the highest average impressions on Company Pages, and Socialinsider's Q2 2026 poll engagement rose to 4.50%. Votes are cheap, though, and LinkedIn's guidance lists reaction-vote style posts under engagement bait. A poll that genuinely asks your audience something you will use is fine. A poll designed only to farm votes is not.
Link posts
Link posts have the lowest engagement rate in Socialinsider's data. The picture on reach is more mixed than the "links are punished" folklore suggests, and we cover it in the next section.
The LinkedIn Algorithm by Account Type and Size
The same post performs differently depending on who publishes it. This is one of the clearest patterns in the data.
Personal profiles
Metricool's 2026 study found personal profiles have a 63% higher average engagement rate than Company Pages and drive significantly more comments, although both get similar average impressions per post. Personal posts benefit from relationship signals that Pages rarely build: people reply to people.
Company Pages
Pages get shared more, according to Metricool, and polls and carousels outperform the image and video posts that make up nearly 75% of what Pages publish. Only 7% of Company Pages in its sample grew enough to move into a higher follower tier. LinkedIn's Pages best practices state that "companies that post weekly see a 2x lift in engagement" and that "Pages with complete information get 30% more weekly views".
Founders and executives
In RevBoss's 2026 benchmark of 28 B2B companies, founder posts earned a median of 520 impressions and company pages 244, a 2.1 times gap, although the report calls the company-page sample small and "directional". The median founder post earned 502 impressions and the 90th percentile 3,334. Those are useful numbers to calibrate against: if your typical post gets 500 impressions, you are normal, not broken. The B2B marketing page shows how teams run founder and Page posting side by side.
By follower count
Socialinsider breaks average impressions down by Page size, and the shape is not linear. Formats that win for small Pages are not always the ones that win for large ones.
| Page followers | Native doc | Multi-image | Video | Text | Poll | Link |
|---|---|---|---|---|---|---|
| 1K to 5K | 525 | 490 | 355 | 300 | 340 | 235 |
| 5K to 10K | 550 | 1,210 | 590 | 425 | 530 | 70 |
| 10K to 50K | 1,150 | 1,850 | 1,420 | 845 | 1,255 | 340 |
| 50K to 100K | 2,600 | 4,320 | 2,390 | 1,695 | 4,845 | 50 |
| 100K to 1M | 1,180 | 3,867 | 2,410 | 4,000 | 9,797 | 2,765 |
Source: Socialinsider LinkedIn benchmarks, average impressions per post by format.
Two things stand out. Multi-image posts are the most reliable format across the middle tiers, and link posts are weak for Pages under 100,000 followers. Small Pages should treat links as an occasional format, not the default.
What LinkedIn Says vs What the Data Shows
A lot of LinkedIn advice is folklore. The table below sets the official position next to the best available independent data, so you can judge each claim for yourself.
| Claim | What LinkedIn says | What studies show | Verdict |
|---|---|---|---|
| Links in posts kill reach | No published link penalty | Metricool 2025: link posts +4.9% impressions; Metricool 2026: personal profiles with links see lower impressions, Pages higher | Depends on account type; not a universal penalty |
| Use lots of hashtags | Nothing official on hashtag counts | Ordinal: 0 hashtags averaged 6,632 impressions, 11 or more averaged 502 | Few or none |
| Pods boost reach | Policies ban agreeing "ahead of time to like or re-share" | Named in the 2026 demotion list | Do not use |
| "Comment YES for the guide" | Engagement bait, may stay in your network | Named in the 2026 newsroom post | Do not use |
| Weekly posting helps | Pages posting weekly see a 2x engagement lift | Socialinsider: Pages averaged 7 image and 4 video posts a month in 2025 | Consistency matters |
| There is one best time | Recency is a content signal | Buffer and Hootsuite reach different answers | Test your own audience |
External links
The link question has the most contradictory evidence. Metricool's 2025 study of 577,180 posts found posts with external links drew 4.9% more impressions and 13.57% more interactions than posts without. Socialinsider's Page data shows link posts with the lowest engagement rate and weak impressions under 100,000 followers. Meanwhile, the LinkedIn engineering work explains why: dwell "after the click" is measured, and the 2020 post describes "click bounces", where someone opens a link and returns within seconds. A link to a page that does not hold attention generates exactly the signal the ranker discounts. Our practical rule: put the link in the post when the link is the point, make the post valuable without the click, and tag it with UTM parameters using the UTM builder so you can see whether the traffic justified it.
Hashtags
Ordinal's analysis found only 26.68% of posts used any hashtag. Posts with hashtags had a higher median engagement rate (2.56% against 1.75%) but much lower average impressions (2,269 against 6,632). Ordinal itself calls narrowed distribution the "most plausible explanation" rather than a tested finding, and it is correlation, not proof. Combined with LLM retrieval that reads topics from the text, there is no longer a good reason to add more than two or three, and many strong creators use none. Sprout Social's January 2026 guide still suggests three as a rule of thumb, which is a reasonable ceiling.
Posting time
Timing advice disagrees because audiences differ. Buffer's analysis of more than 4.8 million posts, published September 2026, puts the strongest window at 3 p.m. to 8 p.m. on weekdays, with Wednesday at 4 p.m. on top, and notes that peak engagement used to sit almost entirely within working hours. Hootsuite's research points to 4 a.m. to 6 a.m. on Tuesdays and Wednesdays. Both cannot be universally right. Recency is an official signal, so timing matters, but the right time is the one your audience is reading and you are free to reply. Our best time to post on LinkedIn guide explains how to find yours.
Scheduling tools
LinkedIn has not published any penalty for posts made through its official API, and none of the ranking documentation above mentions how a post was created. What the 2026 newsroom post does target is "unauthorized third-party tools that create fake conversations", meaning comment automation and engagement tools. A scheduler that publishes your own post through LinkedIn's official API is a different thing from a bot that comments for you.
How a Post Moves Through Its First 48 Hours
LinkedIn does not publish a timetable, and the "golden hour" numbers that circulate online have no official source. What is documented is that new posts are embedded "in near-real time", that the whole nearline process "happens within minutes", according to the engineering post, and that recency is a content signal. Put together, the sequence below is a reasonable model, with the timings as illustrations, not rules.
| Window | What is likely happening | What to do |
|---|---|---|
| First minutes | Post is embedded and becomes retrievable; quality filters apply | Do not delete and repost; fix typos only |
| First hour | Shown to the members most likely to match; early dwell and contributions recorded | Be present to reply with substance |
| Hours 2 to 12 | Retrieved for similar members, including out of network if signals are good | Answer new comments, engage with commenters' posts |
| Days 1 to 2 | Recency weight fades; conversation can keep it alive | Reply to late comments; note results for review |
The practical lesson is to publish when you can be around afterwards. That is a scheduling problem as much as a writing one, which is where a content calendar helps: you write in one sitting and place each post in a slot where you know you will be free.
A Worked Example: A B2B Team Planning a Week
Take Ledgerline, a fictional 12-person company selling reconciliation software to finance teams. It has a Company Page with 6,000 followers and a founder, Priya, with 9,000 followers. Their posts were averaging around 400 impressions and almost no comments. Here is how they reworked a week using what the algorithm documentation and data actually say.
| Day | Account | Format | Topic and hook | Why it fits the ranker |
|---|---|---|---|---|
| Tuesday | Priya | Text | "We closed the books in 3 days instead of 9. Here is the one change that did it." | Specific claim earns the "see more" click and long dwell |
| Wednesday | Page | Multi-image (in the LinkedIn app) | Five-step month-end checklist, one step per image | Swipes build dwell; strongest mid-size Page format in Socialinsider data |
| Thursday | Priya | Single image | Chart of reconciliation errors by source, with her reading of it | Information-carrying image, invites expert comments |
| Friday | Page | Video, 60 seconds, captioned | Controller explains one common audit finding | Video matches its caption; avoids a named demotion pattern |
| Following Monday | Priya | Text with link | Lessons from a customer webinar, link to the recording at the end | Value stands without the click; link tracked with UTMs |
Priya writes all five in one Monday afternoon. Four of them, the text posts, the single image and the captioned video, go onto the Postprism calendar for both her profile and the Ledgerline Page; Postprism publishes text, single-image and video posts to LinkedIn profiles and Company Pages. She blocks thirty minutes after each scheduled slot for replies. After four weeks, she compares comments and impressions per post in LinkedIn's analytics and keeps the formats that earned the most long reads and replies.
A Step-by-Step Method to Learn Your Own Algorithm
General studies tell you what is true on average. Your audience is not average. This method takes about six weeks and gives you answers specific to your account.
- Write down your three topics. The retrieval system matches topics. Pick the two or three subjects you can write about credibly every week, and make sure your headline and About section say the same thing.
- Set a baseline. Export the last 20 posts from LinkedIn's analytics and record impressions, reactions, comments and reposts for each. Calculate your engagement rate per post with the LinkedIn engagement calculator, and note the median rather than the average so one outlier does not mislead you.
- Choose a cadence you can keep. LinkedIn's Pages guidance ties weekly posting to a 2x engagement lift. Three posts a week for a profile and two to four for a Page is a realistic starting point. Our guide to posting frequency goes deeper.
- Batch-write and schedule. Write a week or two of posts in one sitting and place them on a calendar. Bulk scheduling in Postprism lets you load several posts at once and spread them across your chosen slots.
- Test one variable at a time. For two weeks, vary only the format, keeping topic and time similar. For the next two, vary only the time. Changing everything at once tells you nothing.
- Protect the first hour. Schedule posts for times you can reply. Reply to every comment with something new: a detail, a question or a counterpoint.
- Read the right metrics. Comments per thousand impressions and reposts with commentary tell you more than likes. Our LinkedIn metrics guide explains which ones to track and how to read them.
- Keep what works, drop what does not. After six weeks, rank your posts by comments and impressions and look for patterns in topic, format and opening line. Turn the winners into templates for the next month.
If you are building a profile from scratch, our how-to on growing LinkedIn followers pairs well with this method.
Common LinkedIn Algorithm Mistakes, and How to Fix Them
- Asking for comments as a mechanic. "Comment GUIDE and I will DM it" is listed bait. Fix: put the resource in the post or your Featured section, and ask a real question instead.
- Posting about everything. A sequence model learns who you are relevant to from your topics. Fix: stick to two or three themes for at least a quarter.
- Judging posts by likes alone. Long Dwell is a primary target, and reactions are falling platform-wide. Fix: track comments, reposts and impressions together.
- Vague opening lines. If the first two lines look like every other post, members skip, and skips are modelled. Fix: open with a specific number, outcome or claim.
- Pasting video under unrelated text. LinkedIn named mismatched video as a demotion pattern. Fix: write the caption about the video, and caption the video itself for muted viewing.
- Stuffing hashtags. Heavy hashtag use correlates with far lower impressions in Ordinal's data. Fix: use two or three at most, or none.
- Joining pods or using comment tools. Coordinated likes break LinkedIn's Professional Community Policies. Fix: build real relationships by commenting thoughtfully on posts in your niche.
- Recycling the same post every month. Recycled content is on the 2026 list. Fix: update an old idea with a new example, number or angle.
- Promoting without insight. Promotional posts with no value may stay in your network. Fix: lead with what the reader learns, then mention the product.
- Posting when you cannot reply. Early conversation quality is a content signal. Fix: schedule into slots where you are free for half an hour.
- Treating the Company Page as the main channel. Personal profiles earn more engagement in every dataset above. Fix: run founder and employee posts alongside the Page.
- Disappearing for weeks. Relationship and recency signals both decay. Fix: plan a month on a calendar so a busy week does not empty your feed. Our guide to staying consistent on social media has a system for this.
For more on the writing side, see LinkedIn post mistakes.
Where Postprism Fits
Everything above points to the same habit: a few focused posts every week, published when you can reply, on topics you own. The hardest part is keeping it up. Postprism is a scheduler first. You plan LinkedIn alongside five other networks on one calendar, publish text, image and video posts to personal profiles and Company Pages, and choose Public or Connections-only visibility per post. When you share the same idea to X or Bluesky, each network gets its own version, which you can edit yourself or, if you want, have AI adapt to the platform. The LinkedIn scheduler page has the details. If you post open roles every week, our social media scheduler for recruiters page shows how to schedule them to profiles and Pages in one upload.
Schedule Your LinkedIn Posts in Advance
Write a week of LinkedIn posts in one sitting, put each one in a slot where you are free to reply, and let Postprism publish them to your profile or Company Page on time.
Frequently Asked Questions
How does the LinkedIn algorithm work in 2026?
A language-model retrieval system matches each member with posts on topics they care about, then a sequence model that reads their last thousand or so interactions ranks those posts by predicted long dwell and contributions such as likes, comments and shares. Low-quality and bait posts are limited to the author's own network.
What is 360Brew?
360Brew is a 150-billion-parameter research model LinkedIn described in a January 2025 paper as able to handle more than 30 prediction tasks across the platform. The arXiv listing now shows the paper as withdrawn. LinkedIn's March 2026 engineering post describes the production feed as an LLM-based retrieval system plus a Generative Recommender ranker, so treat claims that "360Brew runs the feed" with caution.
Does LinkedIn penalise posts with external links?
LinkedIn has not published a link penalty, and studies disagree. Metricool's 2025 data showed slightly higher impressions for link posts, while Socialinsider's Page data shows link posts with the lowest engagement rate. Include links when they are the point of the post and make the post useful without the click.
How important is dwell time on LinkedIn?
Very. LinkedIn's engineering team has used dwell time in ranking since at least 2020 and added a Long Dwell classifier in 2024. The current Feed SR ranker uses Long Dwell as one of its two main prediction targets.
Do hashtags still help on LinkedIn?
Not for reach. In Ordinal's study, posts without hashtags averaged far more impressions than posts with many. With retrieval now reading topics from the text, two or three relevant hashtags at most is plenty.
Is a personal profile better than a Company Page?
For engagement, yes. Metricool found personal profiles have a 63% higher average engagement rate, and RevBoss found founder posts earned about twice the median impressions of company pages. Most B2B teams do best running both.
How often should I post on LinkedIn?
Consistency beats volume. LinkedIn says Pages that post weekly see a 2x engagement lift. Three posts a week from a personal profile, scheduled for times you can reply, is a sustainable target for most people.
Does scheduling LinkedIn posts hurt reach?
There is no public evidence of that. LinkedIn's 2026 update targets comment automation and tools that fake conversations, not posts published through its official API. Postprism publishes through LinkedIn's official API, and what affects reach is the post, its timing and how you engage afterwards.
