Three quarters of the LinkedIn content that AI assistants cite comes from individual member profiles. Company Pages produce the remaining quarter. That split comes out of a Meltwater study of 9.5 million AI citations across six models, and it quietly invalidates how most B2B companies still allocate organic effort on the platform.
I have spent 22 years buying attention, most of it in paid channels where the feedback loop is fast and unforgiving. Organic social was always the part of the mix I trusted least, because the measurement is soft and the incentives reward applause. AI search changes that arithmetic. When a buyer asks an assistant which vendors to shortlist, the answer gets assembled out of cited sources, and a citation is countable in a way that a company post impression never really was.
The number that should reorganise a content budget
75% of LinkedIn citations came from individual member profiles. Company Pages produced 25%.
Read that as a media buyer rather than as a social media manager. If a channel delivered three quarters of its qualified reach through one inventory type and a quarter through another, and the better-performing inventory was also the cheaper one, the budget would move inside a week. On LinkedIn the opposite happens. The company page gets the design system, the approval chain, the scheduling tool and the agency retainer. The people whose names actually get cited post when they remember to, from their phones, between meetings.
The effect concentrates further by job title. CEOs produced the most-cited content at 8.2%, with founders just behind at 7.5%. That is not a story about charisma. An assistant summarising a market has to attribute a claim to something, and a named practitioner with a track record is a cleaner attribution target than a logo. LinkedIn also received roughly 11 times more citations than Quora over the study window, and its citation share rose 26% in four weeks, so the platform itself is gaining ground while the asset most companies optimise sits on the wrong side of the split.
Recency is a harder filter than authority
48% of successful citations were published within the last three months. Only 12% of cited material was older than a year. For anyone who has spent money on the evergreen-pillar-page theory of content, that is an uncomfortable pair of numbers. The library of durable explainers that ranked well in classic search is close to invisible here. What gets cited is what was said recently about a market that is still moving.
Structure matters more than polish. 92% of top-cited posts featured clear headings, 83% of citations were articles and plain text posts, and 54% of the most-cited content were listicles ranking tools or vendors. Every one of those findings rewards the format a busy practitioner produces naturally and punishes the format a brand team produces expensively: the carousel, the launch graphic, the video with the motion-design intro.
On one B2B account I run, the paid side had been healthy for two years while organic sat flat. Nothing about the content was bad. It was competent, on-brand, scheduled, and posted entirely from the company page. When I looked at which assets had ever been quoted back to me by a prospect, none of them were page posts. Every one was something a named person had written in the first person, usually arguing with a piece of received wisdom.
Volume is not the way through
The obvious response to a citation market is to flood it, and that response is already failing. Over 1 million people clicked LinkedIn’s report option for AI slop between its July 30th launch and a late-August disclosure. That is a platform-scale immune response, built and shipped inside a month, and it exists because the flood already arrived. Any strategy whose first move is more posts per week is competing directly against a filter that a million people have voted to strengthen.
The 72% figure for original content in the same study points the same direction. Cited material is overwhelmingly first-hand rather than rewritten, which is the one quality an automated content pipeline structurally cannot fake.
What I changed
- Budget moved from page production to helping three named people publish two or three posts a week, plus a longer article each month.
- Headings and ranked lists in everything, because the parsing preference is measurable and free to satisfy.
- Anything older than a quarter gets refreshed with new data or retired, rather than kept as a permanent asset.
- The company page keeps exactly two jobs: paid distribution, and looking credible to somebody who lands on it cold.
Where I might be wrong
This is one study, over a four-week window, using one vendor’s tooling to sample six models. Four weeks is short enough that a single platform experiment could move the numbers, and citation share is a moving target that LinkedIn has every commercial reason to influence. If LinkedIn decides company pages should surface more often in AI answers, it can change that split with an indexing decision rather than a product launch, and this argument weakens considerably.
I also hold a bias worth declaring. I have always thought brand-account organic was overfunded relative to what it returns, so a study that says the named individual outperforms the logo is a study I was primed to believe. The honest position is narrower than the headline: the company page is not dead as a business asset, it is dead as an organic citation asset. Those are different claims, and only the second one is supported by these numbers.
What convinces me anyway is that the finding is not really about LinkedIn. Assistants attribute to sources they can name and date. A logo is a weak attribution target in any system built that way, and that constraint is not going to loosen because a brand team would prefer otherwise.