On this page
- The published figures contradict each other, and that is the finding
- An observational study cannot separate a link from the post that carries it
- The Ra-Aha Link Study Audit
- What is known, what is mechanism, and what is folklore
- The first comment workaround rests on weaker evidence than the penalty
- Run the only test that applies to your account
- Price the link in clicks, not in reach
- What moves your reach considerably more than the link does
Four figures circulate for this question and they contradict each other, which is itself the finding. None of the published analyses states its comparison group, holds format constant, or separates the link from the kind of post that tends to carry a link. The direction is broadly agreed and the size is unknown. Audit any claim with five questions, then run a paired test on your own account, because your number is the only one that applies to you.
The published figures contradict each other, and that is the finding
Search this question and you meet four different figures on the same results page. One is a single large percentage reduction in reach for any post containing a link. One is a composite reach score presented on a hundred point scale. One is a range for links placed in the body of a post. One is a much smaller reduction for links moved into the first comment. They cannot all be right, and more usefully, they are not even measuring the same thing.
The table below sets them beside each other by method rather than by number. Read the third column, because the missing information is the same in every case and it is the information that would settle the question.
| The claim as it circulates | What it is measuring | What it never states |
|---|---|---|
| A single large percentage reduction for any post containing an outbound link | Presumably impressions, though the metric is frequently left unnamed | The comparison group, and whether the same authors appear on both sides of it |
| A composite reach score on a hundred point scale | An index built by the publisher from their own weighting of several inputs | The weights, so the score cannot be reproduced, checked, or compared to anything else |
| A range rather than a point estimate for links in the body | A band, which is the more honest presentation of the four | Whether the band expresses variation between accounts or uncertainty in the estimate |
| A much smaller reduction for links placed in the first comment | The residual cost of the standard workaround | Whether the same posts were tested both ways, which is the only design that could answer it |
The four claim shapes in circulation, compared by what they disclose rather than by what they assert.
Restating an unsourced percentage is how it becomes a fact. Every one of these figures is currently travelling from blog to blog with its methodology stripped off, and adding another page that repeats them makes the problem worse rather than better. What is reproduced here is the shape of each claim and the disclosure it is missing, which is the part you can act on.
There is one more thing none of them states. No LinkedIn help page or published engineering note describes a link penalty as a rule, which means every figure in circulation is derived from outside observation of a system nobody outside the company can see. That does not make the observations worthless. It does mean they are estimates of an association, not readings of a setting.
An observational study cannot separate a link from the post that carries it
The central problem is that posting a link is a choice, and the people who make that choice write differently from the people who do not. Any study that gathers posts in the wild and sorts them into link and no link is comparing two populations that differ in several ways at once, and the link is only one of them. This is the flaw that no page on this topic mentions, and it is large enough to account for most of the effect on its own.
- Intent differs. A post carrying a link is usually promoting something, and promotional writing performs differently for reasons unrelated to the presence of a URL
- Format differs. A link in the body renders a preview card, so the comparison is a card post against a plain text post rather than a link against no link
- Copy differs. Link posts commonly end by directing the reader elsewhere instead of finishing a thought, which gives the reader nothing to reply to
- Author mix differs. Accounts that post links regularly skew towards newsletters, publications and company adjacent profiles, whose follower composition is not typical
- Timing differs. Link posts cluster around launches and publication days, which are not neutral moments in an audience's week
The size of the selection problem is easy to underestimate. LinkedIn reports that 89% of B2B marketers use the platform for lead generation, so the population of link posts is heavily weighted towards people whose purpose is to move somebody off the platform and into a funnel. Sorting posts by whether they contain a link therefore sorts them, to a large degree, by whether they were written to sell.
of B2B marketers use LinkedIn for lead generation. That is why the link and no link groups in any observational study are not comparable populations: one of them is disproportionately made of people posting in order to send you somewhere.
LinkedIn, 2026None of this means the effect is imaginary. It means a study of this design can establish that posts containing links reach fewer people, and cannot establish that the link is why. Those are different claims and only one of them tells you what to do on Tuesday morning.
The Ra-Aha Link Study Audit
Five questions decide how much weight any link penalty claim deserves, and they work on the next study as well as on the current four. Ask them in order and stop at the first one the publisher cannot answer, because a claim that fails an early question does not get rescued by a large sample.
Run the audit on the four claims and the result is uncomfortable. The largest sample fails questions two, three and five, which means it can support a statement about association and nothing more. The composite score fails question four outright. The two claims about placement fail question five and lack the paired design that would make them checkable. The strongest available evidence is real evidence of something, and it is not evidence of a penalty.
What is known, what is mechanism, and what is folklore
Sorting the claims into three buckets is more useful than picking a number, because the bucket tells you how confidently to act. Known means multiple independent observations agree on the direction. Mechanism means it follows from how a ranking system of this kind must work, which is a weaker form of evidence than measurement but a stronger form than assertion. Folklore means it has been repeated often enough to sound settled.
| Claim | Status | What would change the status |
|---|---|---|
| Posts containing an outbound link reach fewer people on average | Known in direction, unknown in size, and confounded in cause | A within author study that holds format constant and publishes its comparison group |
| The feed ranks for time spent on the platform, so an exit is a cost | Mechanism, and the most defensible reason a link would reduce distribution | Nothing available to outsiders. This follows from how feed ranking objectives are built |
| LinkedIn applies a fixed percentage reduction to link posts | Folklore. No published document describes such a rule | LinkedIn publishing the ranking treatment, which has not happened |
| Moving the link to the first comment recovers most of the lost reach | Contested, and supported by thinner evidence than the penalty it is meant to solve | A paired test on the same accounts with the same posts, which nobody has published |
| Editing the link in after the first hour avoids the cost | Folklore with a plausible mechanism, because editing changes the text the ranker scored | A public test that distinguishes a substantive edit from a typo fix |
| A link post underperforms because it contains a link | Confounded. Link posts differ from other posts in intent, format, copy and timing | Random assignment of links within one author's own posting schedule |
The status column is what should govern your behaviour, not the confidence with which a claim is stated elsewhere.
The mechanism row deserves the most attention, because it is the row that will still be true when this year's figures are obsolete. A feed optimises for members staying on the platform and returning to it. A post whose purpose is to send a member somewhere else works against that objective, whether or not anybody wrote a rule about links. That reasoning survives every algorithm update and does not depend on a single published number.
- The four figures in circulation for the link penalty are not four measurements of the same thing, they are four different metrics with undisclosed methods, which is why they cannot be reconciled.
- No observational study can separate the effect of a link from the effect of the kind of post that carries one, because posting a link is a choice and the people who make it write differently.
- The mechanism argument is stronger than any of the numbers: a feed that ranks for time spent on the platform has a reason to rank an exit lower.
- A single link post landing below your median proves nothing, because the natural spread between your own posts is wider than the effect anyone claims.
- The honest comparison is not reach against reach, it is reach against clicks, and you cannot make it until you have priced both.
The first comment workaround rests on weaker evidence than the penalty
The workaround is asserted more confidently than the problem it solves, and mostly by the same pages. The argument runs that the ranking system scores the post body, so a link parked in the first comment escapes the scoring while remaining available to readers. That is plausible. It is also unverified, because whether comment content is considered in distribution has never been published, and no public paired test exists in which the same posts were run both ways.
The workaround also carries a cost that nobody prices. A link in the first comment is at least one extra tap away, it is not visible in the feed preview, and on some clients the comment has to be expanded before it appears at all. If your reason for including the link was reach, you have kept the reach. If your reason was clicks, you have traded away a meaningful share of the thing you wanted. The trade only makes sense once you decide which of the two you were buying, which is worked through in choosing a link tactic by what you actually want.
Post the same link with the preview card removed and, weeks later, post a comparable link with the card kept. If the card is doing the damage rather than the URL, the two will differ noticeably. This separates two variables that every published study has treated as one, and it takes you two posts rather than a research budget.
Treat the workaround as a reasonable default rather than a proven technique. It has a plausible mechanism, a low cost when clicks are not the goal, and no evidence against it. That is a fair description of most sensible advice, and it is a long way from the certainty with which it is usually delivered.
Run the only test that applies to your account
Your own paired test will beat every published figure for your purposes, because it holds constant the one variable no study can match, which is you. The design is simple and the discipline is in the sample size, since the natural spread between your own posts is wider than the effect anyone claims.
Same format, same rough length, same posting hour band, same subject territory. You are trying to make the link the only meaningful difference between the two groups, which is exactly what the published studies failed to do.
Link post, then no link post, then link post. Sixteen posts in total, which at two posts a week is two months. Alternating rather than blocking protects you from a good month and a bad month landing on different groups.
Impressions are inflated by repeat serves to the same people, so a post can hold its impression count while reaching fewer humans. Members reached is the cleaner number for this question and it is available on every post.
One post that travels unusually far will drag an average across the threshold on its own. The median of eight is stable enough to be worth reading and it takes no more effort to calculate.
If your posts routinely vary by a factor of three, a twenty percent gap between the two medians is not a finding. Write down your spread before you run the test, so you are not deciding what counts as a result after you have seen it.
Here is the arithmetic on assumed figures, which you should replace with your own recordings. Assume your last twenty posts reached between 400 and 2,000 unique members, with a median of 900 and most posts landing between 600 and 1,300. A single link post reaching 550 sits inside that ordinary range and tells you nothing at all, which is the trap most people fall into after one disappointing post. Now run eight pairs. If the link group median is 640 against a no link median of 900, that is a gap of roughly 29% between two stable numbers and it is worth acting on. If the medians are 840 and 900, the difference is inside your own noise and the honest conclusion is that the link is not your problem.
Eight pairs is a practical floor chosen so that ordinary variation does not dominate, not a threshold derived from a formal power calculation. It is offered as a way to stop drawing conclusions from single posts, which is the actual failure mode. If you want statistical confidence rather than a defensible working judgement, you need considerably more posts than most individuals will publish in a year.
Price the link in clicks, not in reach
Reach against reach is the wrong comparison, because a post without a link cannot produce the outcome the link exists to produce. The honest comparison is reach against clicks, and it can only be made once you have priced both sides in the same currency.
Work it through on assumed numbers you should replace with your own. Assume a post without a link reached 900 unique members and produced 18 profile views. Assume a comparable post with a link reached 620 and produced 11 profile views and 40 outbound clicks. The link cost you 280 members reached and 7 profile views, and it bought 40 clicks. Whether that was a good trade depends entirely on what sits on the other end of the click.
- If the link leads to a page with a next step that converts at, say, one in six, those 40 clicks are roughly 6 people entering something you can follow up on, which is very likely worth 7 profile views
- If the link leads to an article with no next step, those 40 clicks are 40 people who left LinkedIn and are now somewhere you cannot see them
- If your goal was credibility rather than traffic, the profile views were the outcome and the link converted your best asset into a worse one
- If the destination is your own newsletter or booking page, the click has a downstream value you can actually calculate, and the comparison stops being a matter of opinion
There is a third option that this framing surfaces and the link debate obscures. Put the destination on your profile rather than in the post, so the reader arrives through a route that costs you no distribution at all, which is what the Featured block is for and is covered in choosing what belongs in a Featured section. The post then does the job posts are good at, which is making somebody curious about you, and the profile does the job it is good at, which is answering that curiosity without a preview card in the way.
What moves your reach considerably more than the link does
The link is a second order variable that receives first order attention, because it is binary, easy to argue about, and requires no judgement to discuss. Several first order variables sit above it, and any of them will swamp whatever the link is doing to you.
- Who your followers are, and specifically whether the people who followed you for one topic still engage with the topic you write about now
- Whether the post gives a reader something specific enough to disagree with, since a post nobody can argue with produces a thread of polite agreement and stops
- The format, and whether it is a format your particular audience actually consumes rather than the format a benchmark report recommends
- Whether your account is being distributed normally at all, which is a separate diagnosis and not a per post question
- Artificial engagement, which carries a risk profile very different from a URL, and which is examined in the real cost of engagement pods
Content Marketing Institute figures cited in 2026 put LinkedIn at 76% among B2B marketers as the most effective channel for thought leadership, which is worth holding beside the link question. The platform is where people go to establish a view, and a post built around a view will out reach a post built around a redirect regardless of where the URL sits. The link argument is a tactical question that has been promoted to a strategic one.
The reasonable position, stated without the confidence everyone else brings to it, is this. Links probably cost you some reach, the size is unknown and varies more by account than by placement, the first comment is a sensible default when clicks are not the point, and the whole question deserves about ten minutes of your attention rather than the weeks it currently receives.
Questions people ask next
Has LinkedIn ever confirmed that it penalizes external links?
Should I put my link in the first comment instead of the post?
Does editing a link into the post an hour later avoid the effect?
Do links to LinkedIn newsletters or articles count as external links?
How many posts do I need to test this properly on my own account?
Is it the link that costs reach, or the preview card it generates?
Send your profile, keep the aha.
One email with your LinkedIn URL. One specific, actionable insight back, free. If it is obvious, we did not look hard enough.
Email your profile →
