One number should change how most B2B teams run LinkedIn: about 75% of LinkedIn's citations in AI answers come from individual member profiles, and only about 25% from company pages. That is the finding from a 2026 Meltwater study — run with LinkedIn — of roughly 9.5 million AI citations across more than a dozen B2B categories and six major AI platforms, the same study that ranked LinkedIn the second most-cited source in AI answers behind only YouTube. It means the corporate account most brands pour their LinkedIn effort into is the smaller quarter of the opportunity, while the surface that actually gets quoted — the personal profiles of the experts on your team — is usually left to run itself. The study went further: roughly 51% of cited creators had fewer than 10,000 followers, so this is not a reach game; and the cited posts shared a precise format fingerprint — almost all used lists, most had clear headings, and two-thirds carried real data. This guide is about the strategic consequence, not just the stat: why answer engines favor people over pages, what the company page is still good for, the follower myth, the shape of a citable post, and how to actually run an expert-led LinkedIn program across several voices without it collapsing into generic filler or eating your whole week.
Most B2B teams run LinkedIn as a company-page exercise: a marketing team drafts polished posts for the corporate account, and the executives and experts who actually know the subject post rarely, if at all. A 2026 Meltwater study of roughly 9.5 million AI citations — conducted with LinkedIn — says that has the leverage backwards. About 75% of LinkedIn's citations in AI answers came from individual member profiles, and only about 25% from company pages. The surface most brands neglect is the one AI engines quote three times as often as the one they invest in.
This is not a small optimization. As people increasingly ask ChatGPT, Google's AI Mode, Copilot, and the rest instead of running a search and clicking, being the source an answer engine cites is a distribution channel of its own — and on LinkedIn, that channel runs almost entirely through people, not pages. The strategic consequence is uncomfortable but simple: your best LinkedIn AI-visibility asset is not your brand account. It is the personal profiles of the experts on your team, and whether you help them publish or leave them idle.
The study analyzed about 9.5 million AI citations across more than a dozen B2B categories and six major AI platforms — including ChatGPT, Google's AI Mode and AI Overviews, Gemini, Copilot, and Claude — and ranked LinkedIn the second most-cited source of any kind in AI answers, behind only YouTube. That ranking alone reframes LinkedIn from a social feed into one of the highest-authority domains an answer engine reaches for on professional and B2B questions. Inside that authority, the 75/25 individual-versus-company split is the finding with the clearest operational instruction.
One honesty note up front: this was a Meltwater–LinkedIn collaboration, so treat the exact figures as directional and platform-adjacent rather than fully independent. The direction, though, is broadly corroborated — a separate Semrush analysis of roughly 89,000 cited LinkedIn URLs found individual members outweigh company pages on ChatGPT Search and Google AI Mode (about 59% of citations on each), though it found the reverse on Perplexity, where company pages lead. That engine-by-engine split is a real caveat, not a footnote — but across the tools most B2B buyers actually use, individual voices still carry the larger share of LinkedIn's citations, so the strategic instruction survives even as the exact percentages move. For the post-level mechanics of what gets extracted, the companion LinkedIn optimization for AI discovery guide works through the Semrush dataset in detail; this guide is about the individual-versus-page strategy that sits on top of it.
The split is intuitive once you picture what an engine is doing. When a model builds an answer, it is looking for a credible source it can attribute a claim to. A named practitioner writing "here's how we cut render costs 40% on this workload, and the trade-off nobody mentions" is exactly that — first-hand, specific, and attributable to a person with demonstrable expertise. A company page writing "our platform helps teams do more with less" is not a source of a claim; it is marketing, and models have learned to treat it as such. The individual profile carries a perceived authority the brand account structurally cannot.
There is an entity-and-author signal underneath it, too. A well-maintained LinkedIn profile is, to a model, an entity page: a headline and About section that state plainly what this person is an authority on, backed by a consistent stream of posts on that topic. That alignment — the profile claims a domain, and the content demonstrably covers it — is a trust signal the model can use to decide whether to cite the person on a given query. Company pages rarely read as a single coherent expertise; they cover everything the business does, which makes them a weaker match for any specific question.
None of this means deleting the company page. It earns about a quarter of LinkedIn's citations for a reason: brand-, product-, and category-level queries — "what does X's platform do," "who are the vendors in this category" — are exactly the answers a Company Page is the natural source for. The error is not having a page; it is spending 90% of your LinkedIn effort on the surface that earns 25% of the citations. The fix is reallocation: keep the page publishing category and product authority, and put the larger share of your production behind the individual voices that earn the other 75%.
The most liberating number in the study is that about 51% of the cited creators had fewer than 10,000 followers. AI citation is not a reach lottery. An engine does not check your follower count before deciding whether your sentence is the cleanest available answer to a question — it evaluates the text. That decouples AI visibility from the follower-growth grind that dominates feed strategy: a subject expert with a modest audience but specific, well-structured posts can be cited more often than a brand page with a hundred thousand followers publishing generic thought-leadership.
This matters for who you enlist. It means the mid-level engineer, the practitioner, the person with real domain depth but no interest in becoming a LinkedIn influencer is a viable citation source — you do not need to grow them into a 50,000-follower personality first. You need them to publish clear, original, specific content on their lane, consistently. The reach will help human distribution if it comes, but it is not the prerequisite for being the source an answer engine names.
The cited content in Meltwater's sample shared a precise shape, and it is worth memorizing because it is directly actionable. Essentially all of the most-cited posts used bullet points or numbered lists. About 92% had clear H2/H3 structure. Around 75% named specific entities — companies, tools, real things. And roughly 67% included statistics or hard data. Original content beat reshares, and fresh material from the last few months was over-represented. The formats that recurred were decision-led: how-to guides, "how to choose" frameworks, comparisons, and ranked lists.
Read together, that is a single instruction: write answer-shaped content. State a specific claim plainly, structure it so a model can extract one self-contained passage, name real entities, and back it with a number. This is the same specificity that earns AI citations everywhere else, sharpened for LinkedIn's professional context — vague, hedged, could-be-anyone posts score low on the semantic match an engine runs and get passed over, no matter who wrote them. Expertise gets you into the room; structure is what gets you quoted.
Knowing that individuals get cited is easy. Running a program that reliably produces structured, original content from several busy experts, every week, in one recognizable brand, is the hard part — and it is where nearly every attempt stalls. The strategy is not in doubt; the production is.
Start by identifying the people with real domain depth — the founder, the head of product, a few practitioners — and assign each a lane they can own credibly. The goal is a bench of individual voices, each publishing first-hand knowledge in a specific area, so that whenever an engine crawls your category there is fresh, attributable expertise from a named person to find. A single overworked marketing account cannot cover that surface; a coordinated set of expert profiles can.
The other half is rhythm. The data ties citation to a sustained supply of original, on-topic content, which in practice means roughly two to three genuinely useful posts a week per expert — a load few subject-matter experts can carry on top of their actual jobs. Multiply that by several people, all held to one brand voice and the same structured format, and you have a coordination and throughput problem that is exactly why most companies default to the easier, weaker option of one company page. Solving the throughput without letting the content go generic is the whole game — because scaled volume that drifts into filler is precisely what the saturation on LinkedIn and X punishes and what an engine declines to cite.
This is the specific problem Kompozy is built to solve, and it is a production problem, not a writing-assistant one. Kompozy is a full AI content generation and multi-platform publishing engine, and its native shape happens to match what this data rewards: a pool of individual personas — several distinct voices plus a brand identity, one primary — rather than a single account. Each persona carries its own Persona Brief that fixes its voice, its lane, and a banned-phrase list, so a founder, a head of product, and three subject experts can each maintain a real, on-brand publishing stream in parallel. From one source — a talk, a call recording, a rough point of view — it generates net-new text posts, long-form articles, and document carousels per voice, the exact formats the study names, not just repurposed clips.
The part that maps onto this finding specifically is that the cited format is explicit and repeatable — lists, clear headings, named entities, real data — so you can hold every persona to that fingerprint in its brief. Scaling from one expert to five then means more structured, extractable expertise across more individual profiles, instead of more could-be-anyone filler that never gets cited. Autopilot keeps each voice publishing at the two-to-three-a-week cadence the data rewards, to LinkedIn and the other primary social platforms, behind a per-post human review gate — the checkpoint that keeps AI-drafted content specific and original enough to be worth quoting. The company page runs alongside as one more voice in the pool, covering the brand and category queries that make up its share of the citations.
Be clear about the boundary. Kompozy cannot manufacture expertise your people do not have, cannot decide which lane each expert should own, and cannot force any engine to cite you — citation is the model's call, and no tool controls it. What it removes is the production ceiling that otherwise caps an expert-led program at a post or two before the calendar wins, when the data says AI visibility on LinkedIn goes to the team that sustains structured, original, first-hand content across several individual voices. For the news version of this finding, see the Meltwater study writeup; for measuring whether any of it is working, AI visibility measurement covers the scorecard.
Because a person's post reads as first-hand expertise and a company page reads as marketing. Meltwater's 2026 study of ~9.5 million AI citations found about 75% of LinkedIn's citations came from individual profiles, ~25% from company pages. Answer engines are looking for a credible source to attribute a claim to, and a named practitioner explaining how something works is a cleaner, more quotable source than a brand account announcing or selling.
No — reallocate, don't abandon. The company page still anchors brand-, product-, and category-level answers and is a legitimate ~25% of LinkedIn citations. The mistake is spending most of your LinkedIn effort there while the 75% surface — your experts' individual profiles — runs itself. Keep the page publishing category and product authority, and put the larger investment behind a bench of individual expert voices.
Fewer than the reach mindset assumes. About 51% of the creators cited in Meltwater's study had under 10,000 followers, which means AI engines rewarded clarity, specificity, and usefulness over audience size. Followers help human distribution in the feed; they are not the gate for whether a model extracts and cites your text. A specific, well-structured post from a mid-sized account can outperform a large brand page.
That coordination is the real difficulty of an expert-led program, and it is where most attempts break. The workable approach is a governing brief per person that fixes each expert's voice, lane, and a banned-phrase list, so several people can publish in parallel and still read as one coherent brand. Tools like Kompozy run a persona pool that does exactly this — many individual voices plus the company page, each governed, published on a schedule behind a review gate.
Kompozy is an AI content generation and multi-platform publishing engine built for the exact shape this data rewards: a bench of individual expert voices producing structured, original, decision-led content at cadence. It runs an AI Influencer persona pool — several personas plus a brand identity, each with its own Persona Brief — and from one source generates text posts, long-form articles, and document carousels per voice, formatted as lists with clear headings and real data, then schedules them to LinkedIn and other platforms behind a per-post review gate.
Meltwater's 2026 study of about 9.5 million AI citations, run with LinkedIn, found roughly 75% of LinkedIn's citations in AI answers come from individual member profiles and only ~25% from company pages — and about 51% from creators under 10,000 followers. The lesson for B2B: AI visibility on LinkedIn is won by arming internal experts to publish structured, original, decision-led content, not by polishing one corporate page.
Get started → · ← All guides · Compare Kompozy vs other tools