Meltwater’s LinkedIn citation study: what the format numbers mean
The May study highlights useful B2B content patterns. Its format percentages describe selected cited content, not your probability of earning a citation.
A chart showing that 54% of highly cited content uses a particular format does not mean that publishing in that format gives your next article a 54% chance of being cited. That distinction matters when reading Meltwater’s research on LinkedIn and AI search.
The study offers useful ideas for B2B publishers: answer a concrete buying question, show the basis for a recommendation, and make an expert’s reasoning easy to follow. It also leaves important questions about the sample and comparisons unanswered. Use it to choose a publishing experiment, then measure what happens in your own topic.
What the LinkedIn and Meltwater research covers
LinkedIn’s article, dated May 21, 2026, describes a four-week B2B study spanning six AI search surfaces. Its headline refers to 9.5 million citations. It reports that 75% of LinkedIn citations came from individual members and that 48% of cited content was less than three months old.
Those observations concern the material found in the study. They do not establish a universal advantage for every personal account or a required publication schedule. This is an analysis of a May report, rather than a newly released November study.
How to read the five content-format percentages
The graphic in this article shows the following figures. Meltwater’s public study summary describes these formats as appearing among top-cited LinkedIn content.
| Content format | Reported percentage |
|---|---|
| Best-of lists | 54% |
| Side-by-side comparisons | 50% |
| Guides to choosing an option | 33% |
| Educational explainers | 17% |
| Thought leadership with data | 8% |
Adding the five values gives 162%. They therefore cannot be interpreted as mutually exclusive shares of a single total. Overlapping labels are a plausible explanation: one article can list products, compare them, and explain how to choose. The public text does not fully specify the format-coding rules or the exact denominator for this graphic, so overlap remains an interpretation rather than a verified reconstruction.
Meltwater separately describes an examination of the 24 most-cited LinkedIn articles for structural features. That small, selected group should not be confused with the headline-scale citation collection. Nor should you automatically assign its size to every other chart.
The missing comparison is the content that was available but never cited. If comparison articles are common in the underlying collection, they might also be common among citations without having a higher citation rate. A format’s prevalence among selected winners cannot answer that question on its own.
A large citation share can hide a smaller success rate
Consider a fictional example involving 100 distinct cited posts. Suppose 75 belong to individuals and 25 belong to companies. Now suppose the eligible collection contains 750 individual posts and 50 company posts.
In that invented collection, 10% of individual posts were cited, compared with 50% of company posts. Individuals supplied most cited posts, while company posts had a higher citation rate. These figures are not Meltwater results; the example shows how the available citation population changes interpretation.
The same problem affects freshness claims. A large proportion of recent content among citations could reflect the supply of recent material, the questions sampled, a preference for recency, or several factors together. A comparison with eligible older content would help separate those explanations.
Build a comparison a reader can actually use
For a publisher choosing its next LinkedIn article, the useful question is whether the proposed structure helps someone make a decision.
Imagine an SEO consultant writing about choosing a rank tracker for a multilingual site. A useful article would name the languages and markets involved, explain the position definition being compared, identify the source and date of each price, and describe what remains untested. Its recommendation would follow from those requirements.
A numbered list can help organize that information. A comparison can expose differences. A short selection guide can explain which difference matters to which reader. All three can belong in one article when the question warrants them.
Adding those elements without evidence would produce a more elaborate page with the same weak answer. Choose the format after defining the decision and collecting the material needed to support it. Our evidence-led publishing guide provides a practical starting point.
What remains uncertain in the public methodology
The two public summaries disagree on some details. LinkedIn’s methodology describes 9.5 million responses, despite its citation-based headline. It names Gemini 2.5 Pro; Meltwater’s page names Gemini 3.5 Pro. These differences prevent a precise reconstruction of the collection from those pages alone.
The full-report page requests a form submission. This article relies on the accessible summaries and supplied graphic; SearchEngineAnswer did not obtain the gated report or inspect underlying answer records. Meltwater sells the measurement product used, and LinkedIn benefits commercially from publishing on its platform. Read the findings as commercially interested observational research.
Before applying the results broadly, request the collection dates, prompt list, language and geography, model versions, repeated-run procedure, and content-selection rules. Ask which charts count citation events, unique URLs, posts, or responses. Each unit answers a different question.
A November snapshot shows why the search surface matters
The Ahrefs UGC visibility benchmark, updated November 8, reports that LinkedIn appeared in 1.37% of 8.64 million tracked Google AI Overview responses collected November 7. When Ahrefs combined AI Overviews and AI Mode, LinkedIn appeared in 1.10% of 13.33 million tracked responses. The combined figure weights each surface by the number of responses collected.
Ahrefs labels these figures “found-in” percentages, with tracked responses as the denominator. Meltwater’s figures describe a different sample and use different units. Read each figure alongside its source, metric, and collection date before comparing reports or judging a content format.
Measure the next article separately from the study
Choose a small set of relevant questions before publishing. Record the wording and collection conditions, repeat observations, and preserve the actual cited URLs. Distinguish a citation of the LinkedIn post from a citation of your website, and record brand mentions separately from either.
Track visits and useful inquiries independently. A cited article can support an answer without producing a measurable website visit. The AI visibility measurement crosswalk explains those boundaries, while our citation-stability analysis explains why one missing citation deserves a repeat check.
For your own website, Google’s AI search guidance says established SEO practices remain relevant, and no special optimization is required for AI Overviews or AI Mode. That guidance applies to Google’s features; it does not validate a preferred LinkedIn format across every answer engine.
Use the Meltwater chart as a source of editorial questions. Pick a reader problem, publish a well-supported answer, and compare repeated observations with the baseline you recorded. A change in citation visibility is worth investigating; attributing it to the format requires stronger evidence.
Keep learning
Continue this topic
Audit citations
Locate the failing answer layer
Compare evidence
Separate mentions, citations and visits
Measure referrals
Check AI search visits in GA4
AEO & AI Search
Ask a question or join the discussion