Do saves and shares help LinkedIn reach? Tracing the claims to source

Shares are a documented LinkedIn ranking signal. Saves are not, and the widely repeated claim that a save is worth five times a like does not survive being checked. LinkedIn has published four descriptions of what its feed models predict, between 2020 and 2026, and they name click, skip, long dwell, like, comment and share. Saves appear in none of them. When we followed the citations behind the strongest version of the save claim, two of them returned 404. Our own dataset has no saves field and no shares field, so we will not be adding a number to the pile. What follows is what is documented, what is asserted, and which is which.

Do saves and shares help LinkedIn reach?

Split the question, because the two have completely different evidence behind them.

Shares: documented. LinkedIn's February 2026 feed ranking paper names two optimisation targets. One is long dwell. The other is contribution, which the paper defines as likes, comments or shares. Its March 2026 engineering post splits the model into “passive tasks (click, skip, long-dwell) and active tasks (like, comment, share)”. The 2024 LiRank paper uses the same split. A share is predicted. That is as close to confirmation as an outsider gets.

Saves: absent.Not contradicted, not dismissed. Simply never mentioned, in any of the four published descriptions, across six years. LinkedIn's post analytics help page defines engagement rate as “the ratio of interactions per impressions on your post” where “interactions include clicks, reactions, comments, and shares”. Saves are not in that list either.

Absence of evidence is not evidence of absence, and we want to be careful about that. It is entirely possible that saves feed ranking and LinkedIn has simply never written about it. What is not reasonable is to publish a precise multiplier for a signal that no public document has ever confirmed exists in the model.

Where does “a save is worth 5x a like” come from?

We traced it as far back as we could. Here is the chain, in the order the claims get stronger and the sourcing gets weaker.

Notice the direction of travel. The original claim is about reach and comes with a name attached. One hop later it has become an algorithmic weight, gained a second multiplier against comments, and lost its attribution. Two hops later it has acquired footnotes to documents that do not exist.

What happened when we followed the citations?

This is the part we think is worth publishing, because almost nobody clicks through.

The OmniCreator page attributes its 2.1x distribution figure to a LinkedIn engineering post published in March 2026, at engineering.linkedin.com/blog/2026/save-rates-authority-signal. We requested it. The server returned HTTP 404 Not Found. There is no such post.

The same page attributes its 8% save rate threshold to a Rival IQ report at blog.rivaliq.com/save-rate-benchmarks-2026/. That host redirects to Rival IQ's current domain, and the destination also returns HTTP 404. Rival IQ has never published a save rate benchmark report that we can find. Its actual LinkedIn benchmark report covers 58,000 posts and reports engagement rate, impressions per follower and posting frequency. It contains no save data.

We are not claiming the author knowingly invented them. It does not much matter. The effect on a reader is the same: two authoritative-looking citations, neither of which leads anywhere, supporting a number that then gets quoted onward as though it came from LinkedIn.

If you take one habit from this page, make it this one. When a LinkedIn statistic looks precise and important, click the citation. In this subject area it fails more often than it holds, which we also found when tracing the claim that organic reach fell 50% and when auditing what is actually known about AI content detection.

What has LinkedIn published about what the feed predicts?

Four documents, six years, one consistent vocabulary. Here they are with the actions each one names.

There is an indirect route by which saving could still matter, and it is worth stating because it is the strongest steelman available. Saving a post takes time on the post. Time on the post is dwell time, and long dwell is one of the two headline objectives in the 2026 model. A save might therefore feed ranking through the dwell channel without ever being a separate signal.

That is a plausible mechanism and it is also unquantified. It gives you no multiplier. If anything, it argues that the thing to optimise is whatever holds attention, which we went through in what LinkedIn has published about dwell time.

What is a LinkedIn save, and who can see it?

A save bookmarks a post into your own list for later. That much is uncontroversial. Almost everything else about the feature's visibility is contested by the pages that discuss it, which is itself informative.

Some guides state flatly that saves are private, that only the saver sees their list and that the author is never notified. Other guides published in the same period state that save counts became visible to authors. Those two claims cannot both be right, and neither camp cites LinkedIn.

What we can verify is the one document that would settle it. LinkedIn's post analytics help page lists what a creator can see: impressions, described as “the number of times your post was shown on LinkedIn”, members reached, described as “the number of distinct members and Pages that saw your post”, and an engagement rate built from “clicks, reactions, comments, and shares”. No saves metric appears anywhere in that definition set.

If your analytics do show a saves number, treat this page as out of date on that point and the rest of it as unchanged, because a visible counter would still not tell you what saves are worth in ranking.

What else does the saves literature claim?

The five-times figure is not the only unmethodologised number in circulation. AuthoredUp's algorithm guide, which is the most careful of the sources we traced and does publish dataset sizes elsewhere on the page, also states that fewer than 3% of posts reach a meaningful save level, that a saved post leads to a 130% higher chance someone follows you, and that creators whose posts get saved consistently grow their audience three times faster.

Each of those is a specific, checkable-sounding number. None of them arrives with a sample size, a definition of “consistently”, a control group, or a date range. The same page publishes a disclaimer worth quoting, because it is more honest than most of what gets built on top of it: “The insights are based on independent research by AuthoredUp and Just Connecting, which are not affiliated with, endorsed by, or officially connected to LinkedIn Corporation.”

That is the correct framing and we would apply it to ourselves too. Independent research on a platform you do not control is inference. It can be good inference. It is not a specification document, and a multiplier stated to one decimal place implies a level of access nobody outside LinkedIn has.

Is a repost the same as a share?

For ranking purposes, LinkedIn's published material does not distinguish them. Its models name “share” as one of the active tasks and one of the three contribution types. There is no published treatment of a bare repost against a repost with commentary, despite the confident hierarchies you will find asserting that one outranks the other.

Our dataset cannot settle it either, because it carries neither. What we can say is that the two behave differently for the person doing it: a repost with commentary creates a new post with its own hook, which then competes on the same terms as any other post, while a bare repost does not. That is a content argument rather than an algorithm argument, and it is the one we find more useful. We went through the trade-offs in what happens when you repost on LinkedIn.

How much is a share actually worth?

Documented as an input, undocumented as a magnitude. LinkedIn names shares as part of contribution and does not publish how the prediction is weighted against long dwell or against the other contribution types. Anybody quoting a share multiplier is doing the same thing the save pages are doing.

What we can say comes from measurement rather than from LinkedIn.

• Shares are getting rarer. Metricool measured them down 10% year over year across 673,658 posts from 63,108 accounts, alongside likes down 13% and comments down 17%.
• Shares are uncommon in absolute terms. Metricool measured company pages averaging 1.20 shares per post against 12.48 likes and 0.48 comments.
• Account type changes the mix. Metricool found company pages receive more shares while personal profiles drive significantly more comments.

One reading of that last point matters for anyone choosing where to publish. If you post from a personal profile, shares are not your main lever regardless of their weight, because your audience is not the group that shares. Comments are. We compared the two account types in more detail in our impressions benchmark by follower tier.

What can our own dataset say about saves and shares?

Nothing, and we would rather say so in a heading than bury it. Our dataset covers 12,988 English posts from 65 creators with at least 1,000 followers. The fields it carries are reactions, comments, follower count, media type and post text up to the see-more fold. There is no saves column and no shares column. Any figure we published about either would be invented.

What the data does support is a claim about the interaction type that is documented, visible and measurable: comments.

Reactions roughly triple between the bottom half and the top decile. Comments increase more than fourteen-fold. That asymmetry is why we weight comments at four times reactions when we score a post, and it is the reason we would spend a call to action on a comment rather than on a save. One of those two is documented as a ranking input, visible in your dashboard, and strongly associated with the tail of the distribution. The other is none of those things.

Why does the save folklore keep spreading?

Three reasons, and understanding them is a decent defence against the next one.

1. Saves are invisible, so the claim is unfalsifiable. You cannot open your analytics and check whether your saves outperformed your likes. Advice built on a metric nobody can see can never be contradicted by a dashboard.
2. It has a satisfying story attached. A save feels like a stronger vote than a like, so a multiplier feels obviously right. Plausibility is doing the work that evidence should be doing.
3. The number is portable.“5x” survives copying in a way that a methodology does not. Each hop drops a qualifier, and after three hops a vendor's internal observation has become a law of the platform.

The same mechanism produced most of the confident numbers in this niche. It is why we go looking for the sample size and the date range before we quote anything, a discipline we applied to the whole benchmark literature in our breakdown of average engagement rates.

How would you test whether saves help on your own account?

You mostly cannot, and it is worth understanding why before you spend a month trying.

A proper test needs the treatment to be observable. You would need to know how many saves each post received, and to compare posts that are otherwise similar. LinkedIn does not expose a saves count in the analytics definitions it publishes, which removes the first requirement. Without it, any experiment you run is measuring the call to action rather than the save.

Here is the honest version of what you can do.

1. Run a call-to-action test, not a saves test. Publish ten comparable posts asking readers to save, and ten asking a specific question. Compare impressions and comments across the two sets.
2. Measure the opportunity cost. The interesting result is not whether the save posts did well. It is whether they did better than the question posts, given that Metricool measured posts with questions receiving 77% more comments.
3. Hold format constant. Save calls to action cluster on carousels, and carousels outperform on every published format table. If you do not control for that you will credit the save request for the format.
4. Use medians and at least ten posts a side. One breakout will dominate an average and tell you nothing.
5. Accept the result is about the ask. Whatever you learn, you have learned which call to action works for your audience, which is genuinely useful and is not the same as learning what saves are worth. Our guide to LinkedIn calls to action covers the patterns worth testing.

What would change our mind?

Any one of these would move the save claim from folklore to evidence, and we would update this page.

• A LinkedIn engineering post or paper naming saves among the modelled objectives, in the same way long dwell and contribution are named.
• A saves metric appearing in LinkedIn's published analytics definitions, which would at least make outside measurement possible.
• A vendor publishing a save study with a sample size, a date range, a definition and a control, rather than a multiplier.

Until one of those exists, the correct posture is not “saves do nothing”. It is “nobody knows, including the people telling you they do”.

What are the limits of this page?

• We are arguing from absence. LinkedIn not publishing something is weak evidence. It is stronger than a 404, which is what the alternative rests on, but it is not proof.
• LinkedIn publishes architecture, not weights. Even for shares, which are documented, no public source states how much a share moves distribution.
• Our dataset has no saves or shares. Every claim we make from it concerns reactions and comments only.
• Our data is correlational and comes from 65 established creators above 1,000 followers, with lifetime-cumulative engagement counts at scrape time and post ages that vary. Text findings describe the visible hook.
• Product surfaces change. If LinkedIn ships a public saves counter after this was written, the measurement picture changes and this page will need revising.

Do saves and shares help LinkedIn reach? Restated

Shares are named in LinkedIn's own published feed models as part of contribution, so they are an input. How much of one, nobody outside LinkedIn can say. Saves are not named in any of the four published descriptions from 2020 to 2026, and are not in LinkedIn's analytics definitions either.

The five-times-a-like figure traces to a vendor blog with no method, spreads without attribution, and in its most cited form rests on two citations that both return 404. If you are choosing what to ask readers to do, ask for the thing that is documented as an input, visible in your dashboard and most strongly associated with the tail of the distribution: a comment. The evidence for that is in our study of 12,988 LinkedIn posts, where the median top-decile post earns 72 of them and the median bottom-half post earns 5.