AI tools helped gather research and write the first draft of this article. I edited it and checked every figure and quote against the page it comes from, listed at the end. The cover illustration is AI-generated. Details at the end.
A study of almost 866,000 LinkedIn posts found something more useful than the usual argument about links. When a post's only visual was the link preview card, it got 35% fewer impressions than the same author's posts without a link. When the address sat in the text of an ordinary text post, the gap was 2%, which the authors call "no measurable difference" [2].
So the link itself was not the main problem in that data. The problem was a post that is nothing but a pointer: a headline, a preview card and a request to go somewhere else.
A LinkedIn post that carries an article has to work as a small piece of writing in its own right. It should give the reader one useful part of the article before the link, and say honestly what else the full piece contains. This article shows how to build such a post, and why the evidence points that way.
This is the first part of a new series, Platforms & Metrics. It covers how one article is adapted for LinkedIn, X and other channels, and then how to measure the results on one scale, so that a like on one platform is not mistaken for a reader on another.
What LinkedIn says it measures
LinkedIn does not publish the weights of its feed. Its engineers do describe the families of signals. In an engineering post from 1 October 2024, they separate "passive behaviors such as clicks, dwell time" from "active contribution behaviors such as comments, reshares" [1].
Dwell time, they write, "refers to the duration users spend on content or posts" [1]. It is more plentiful than likes or comments, and the team calls it "a critical indicator of not only content quality but also relevance" [1]. Two details matter for anyone adapting an article:
- Quick scrolling counts against a post. Updates viewed for less than a threshold are classed as "skipped", and the probability of a skip was added to the ranking framework [1].
- Long reading is compared within a content type. The team notes that "a dwell time of 30 seconds could be considered 'long' for an image but not for a video", and normalised dwell time by type so that the model would not simply favour long formats [1].
None of this gives a formula. It does describe what a post is competing for: a reason to stop scrolling, a reason to keep reading, and something worth replying to. A headline and a preview card offer very little of any of the three.
The link question, traced to the data
Ask about links on LinkedIn and you will meet a confident percentage. The best-known is 60%. A Forbes column published on 30 July 2026 carries the headline "The LinkedIn Link Penalty Cutting Your Reach By 60%" [3]. The research behind this article followed the figure back through blogs and Reddit and found no dataset behind it [6]. The column itself could not be read for this article, because Forbes blocks automated access; the headline, author and date were confirmed through a secondary page [3].
The studies that do publish a method give smaller and very different numbers:
- MagicPost compared each post that had a link with the same author's posts without one, published in the same quarter: 124,893 link posts from 9,931 authors, out of 865,980 posts from 15,719 authors between July 2024 and July 2026 [2]. Without adjustments, link posts got 17% fewer impressions. Comparing posts of the same type gave 16%. Comparing posts that were equally "salesy" gave 11% [2].
- By format, the gap was 35% when the preview card was the only visual, 5% for an address in an image post, 2% in a text post and 1% in a carousel or video post [2]. A link in the first comment did no better than a link in the post: 2% fewer impressions, again "no measurable difference", across 1,824 authors who used both [2].
- Ordinal, a LinkedIn tool vendor, puts the gap at 26.5% and says it grew from 5% in 2023 to 42% in 2025. It reports that "company pages face a real reach penalty from external links while personal profiles show almost none" [4]. The same page cites 36% from another vendor, LinkPost [4].
These numbers come from vendors' own customers, measured in different ways. MagicPost lists its own limits: it cannot see how LinkedIn ranks posts, so it reports gaps rather than causes; the "salesy" rating was made by an AI model; and the data covers only accounts it tracks [2]. Ordinal notes that LinkedIn "has never confirmed a deliberate downranking mechanism for outbound links" [4].
What LinkedIn people have said points the same way as the format data. Asked in August 2025 whether a link affects distribution, Rishi Jobanputra, a senior director of product for the LinkedIn feed, gave one condition, as quoted by LinkedScore: "make sure your post gives enough context that it could stand alone without the link" [5]. The original post was not opened for this article; the quote is taken from that secondary page.
Put together, the honest reading is narrow. A link does not doom a post. A post that is only a link probably does worse, and the way to avoid that is the same whatever the true size of the effect: give the post enough content to stand alone.
A post is an adapter, not a trailer
A trailer withholds. It shows a little and asks you to go elsewhere for the rest. An adapter takes the same current and changes its shape so that it fits a different socket.
For LinkedIn, that means the post delivers one complete, useful idea from the article in a form that suits the feed: short paragraphs, one example, one check the reader can apply today. The reader who never clicks still leaves with something. The reader who wants the method, the sources and the limits has a clear reason to open the original.
This also changes what counts as success. The goal of a post that carries an article is a reader who opens the article and reads it, not only a large impression count. More on how to measure that below.
Five parts of a post that stands on its own
- Open with one concrete finding, not the title. A number, a result or a specific situation from the article. The first lines are what a reader sees before deciding whether to stop.
- Develop one piece of evidence. Say where it comes from in a few words: who measured it, how many cases, what year. One well-explained example does more than three listed in passing.
- Give the reader a check or a decision. A short list of questions, a test, a thing to do differently. This is the part that has value even without a click.
- Say what the full article adds. The method, the other studies, the limits of the evidence. Make a promise the article actually keeps.
- Put the link in the text, with a plain line before it. On Dapp.Expert's author posts this is always "Read it on Dapp.Expert:" followed by the address on the next line. Use your own image rather than leaving the preview card as the only visual, since that was the format with the largest gap in the MagicPost data [2].
On format beyond the single post: an analysis by the analytics company Socialinsider of about one million posts from 9,000 company pages found that carousels and document posts drew the most engagement on average, followed by video [7]. It did not include personal profiles, so it says little about an author's own account. A document version of an article is worth testing; it is not a proven upgrade.
Formatting is part of the adaptation
LinkedIn does not render Markdown. Asterisks, hash signs and underscores appear as typed. Paragraph spacing has to be visible in the post itself, and lists work best as short lines starting with a simple bullet character.
One practical lesson from preparing the posts for this site: when the text was inserted into LinkedIn's editor by an automated tool, the blank lines between paragraphs were lost and the paragraphs ran together. The fix was to paste the text with an explicit empty paragraph between blocks, then reload and read the saved post again, rather than trusting the editor's preview. Whatever tool you use, check the live post, not the draft.
A worked example
Take a recent article on this site about social proof, which reports a randomised experiment where one extra up-vote on a new comment made the next reader 32% more likely to up-vote and left final ratings 25% higher on average.
The pointer version below is invented for contrast.
New article: When Social Proof Makes a Bad Decision Easier. Why do we follow the crowd? Link below.
It names the topic and offers nothing to read, check or answer.
The adapter version, from the actual LinkedIn draft for that article, opens with the experiment and its two results and names the study. It then explains in three bullets why online counts often fail the conditions that make social proof useful: they snowball, they can be manufactured, and they may answer a different question. It ends with one check, "Before a number decides for you, cover it and ask what you would decide without it", and only then the line "Read it on Dapp.Expert:" with the address.
A reader who stops there has a finding, a reason and a habit. A reader who clicks finds four more studies, a regulator's rule, a fictional worked case and the limits of the evidence. The post and the article do different jobs, and neither repeats the other.
What to measure afterwards
If the goal is reading, impressions are only the first step. MagicPost notes that LinkedIn has reported link clicks only since June 2026, and does not count clicks on links placed in comments [2]. Both are reasons to measure on your own site as well.
The simplest method is a tagged link. Dapp.Expert's author posts use a standard form: the verified article address with utm_source=linkedin, utm_medium=social and utm_campaign set to the article's identifier. The site's analytics can then show how many visits came from the post and, more usefully, whether those visitors stayed and read. Later parts of this series look at why view counts from different platforms cannot simply be added up, and at what an author's own dashboard should show.
A short check before you post
- Does the first line state a finding, not just the title?
- Is there one piece of evidence with its source?
- Would a reader who never clicks still leave with something useful?
- Does the post promise only what the article delivers?
- Is the link in the text, after a plain line, with your own image?
- Did you read the live post after publishing?
What this article cannot tell you
LinkedIn does not publish how its feed weighs these signals, and the link studies here are vendor analyses of their own customers, with different methods and different results. None of them measures how many readers a post sends to an article, which is the outcome that matters most here. The 60% figure is described as untraced because no dataset was found behind it, not because it has been disproved. The Jobanputra quote is taken from a secondary page; the original post was not opened. The format data from Socialinsider covers company pages, not personal profiles. Nothing here reports results from Dapp.Expert's own LinkedIn posts; those have not been measured in a way that allows a claim.
How this article was made
| Task | What the AI did | What a person did | How it was checked |
|---|---|---|---|
| Research | Collected platform documentation and link studies into research notes | Chose the question and the sources to rely on | Figures and quotes checked against the pages listed below [1]–[7]; secondary quotes are marked as such |
| Draft | Wrote the first draft from the notes | Edited the text, structure and worked example | A separate review compared figures and claims with the sources |
| Images | Generated the cover photograph-style illustration | Chose the illustration; designed the explanatory diagrams | Labelled as AI-generated in the disclosure line and alt text |
| Not done by AI | — | The decision to publish and responsibility for the text | — |
Conflicts of interest: none. The author has no commercial relationship with LinkedIn or with the companies whose studies are cited.
Responsible for this article: Dmitriy Lastovka, author. Corrections: contact form.
Sources
Labels: official (the platform's own page), study (a vendor's published analysis, read on its own page), secondary (a quote or figure reported by another page).
- Fengyu Zhang, Mohit Kothari, Birjodh Singh Tiwana, "Leveraging Dwell Time to Improve Member Experiences on the LinkedIn Feed", LinkedIn Engineering (1 October 2024). https://www.linkedin.com/blog/engineering/feed/leveraging-dwell-time-to-improve-member-experiences-on-the-linkedin-feed Official.
- Naïlé Titah, MagicPost, study of external links and LinkedIn reach (29 October 2026). https://magicpost.in/blog/linkedin-external-links-reach Study.
- Jodie Cook, "The LinkedIn Link Penalty Cutting Your Reach By 60%", Forbes (30 July 2026), headline, author and date as reported by LinkedScore [5]. https://www.forbes.com/sites/jodiecook/2026/07/30/the-linkedin-link-penalty-cutting-your-reach-by-60/ Secondary; column body not read.
- Ordinal, "LinkedIn link penalty study". https://www.tryordinal.com/blog/linkedin-link-penalty-study Study.
- LinkedScore, "Does LinkedIn penalize links?", quoting Rishi Jobanputra's post of 14 August 2025. https://www.linkedscore.app/blog/does-linkedin-penalize-links Secondary.
- Dapp.Expert research notes on platform adapters (5 November 2026), tracing the 60% figure through a Reddit snapshot of 19 February 2026 and later reprints. Author's research; not a published source.
- Andrew Hutchinson, Social Media Today, on Socialinsider's analysis of about one million LinkedIn company page posts (14 April 2025), via Yahoo Finance. https://finance.yahoo.com/news/report-looks-best-performing-linkedin-200014407.html Secondary.










