AI Answer Engines Quote People, Not Company Pages

The LinkedIn account your company spends the most on is the one AI answer engines quote least. Individual posts by named people drew 32.8% of the social citations in a new audit. Company pages drew 6.7%. That’s close to a five-to-one gap pointing the opposite way from where most B2B social budget goes.

The audit is Oktopost’s, published August 26: 63 B2B companies, 27,181 total citations, of which 2,720 came from social sources. Of the 385 citations where an engine quoted a named individual, LinkedIn was the source 285 times. Daniel Kushner, Oktopost’s co-founder and CEO, wrote it up:

“When those engines quote a named person, not a company, not a forum thread, a person, that person published on LinkedIn 74% of the time.”

Medium was next at 12.5%, Substack at 4.9%.

Read who paid for this before you act on it

Oktopost sells employee advocacy software. Its finding is that companies should get more employees publishing under their own names, which is the thing its product does. That’s as direct a commercial interest as this beat produces, and it belongs at the top of the piece rather than in a footnote.

To Oktopost’s credit, the disclosure is in the post rather than extracted from it:

“It’s our own audit, not a published research study, so treat the numbers as a snapshot of what we found.”

Two further gaps you should know about. The piece never names which answer engines it audited, so ChatGPT, Perplexity and Google’s AI Mode are folded into one number despite behaving differently. And the 385-citation base for the headline 74% is small, which is fine for a directional read and thin for a strategy deck.

So treat the direction as credible and the precision as provisional. That’s the same posture we took when three trackers disagreed about Reddit’s citation collapse, and for the same reason: the people measuring this all sell something downstream of the answer.

The direction holds up against an independent source

The useful test for a vendor number is whether anyone unaffiliated saw the same shape. Here, roughly.

Promptwatch, which sells visibility monitoring rather than advocacy software, tracks which LinkedIn page types get cited and finds Pulse articles and individual posts dominating across engines, with company pages well behind on every one except ChatGPT. Its window is older and its own methodology is thinly documented, so it’s corroboration rather than proof.

Two vendors with different products and different incentives found company pages losing to individual accounts. That’s worth more than either number alone.

Why an engine would prefer a person

The mechanism is duller than the finding, and it matters more.

A model generating an answer needs something to attach a claim to. A name, a title, a company, a date. That’s what makes a sentence citable rather than merely true. A company page speaks in the institutional voice, which is exactly the voice with no attributable claimant inside it, and the same page is also where the marketing copy lives, which reads as promotional to a system trying to answer a question.

An individual post supplies the attribution for free. It arrives with a person, a role and an employer already attached.

This is the same underlying pressure we found when answer engines started rewarding specificity over volume. The engine isn’t rewarding humans over brands out of preference. It’s resolving a sourcing problem, and a named human resolves it where a logo can’t.

The counter-case: 58.9% of it is Reddit, and you don’t control that

The number that should temper all of this is the one the headline skips.

Community discussion, mostly Reddit, accounted for 58.9% of the social citations. Individual posts at 32.8% and company pages at 6.7% are dividing up what’s left. So the largest single social source of AI citations is a place where you don’t post, can’t control the framing, and where an employee arriving to advocate is a liability rather than an asset.

And Reddit’s own share is unstable. Semrush reported on August 26 that Promptwatch measured Reddit falling from 3.8% of ChatGPT citations to 0.5% in under a month. Those two numbers don’t contradict each other, because the denominators differ: Oktopost’s 58.9% is a share of social citations across unnamed engines, Promptwatch’s 0.5% is a share of all citations in ChatGPT alone. But it does mean the biggest slice in this audit is the one moving fastest, in a direction nobody has explained.

Read that way, the whole LinkedIn finding is a fight over the second-largest slice. An advocacy program can move 32.8%. Nothing in your control moves the 58.9%, and pretending otherwise is how teams end up astroturfing a forum that punishes it.

Which is the honest shape of the opportunity: real, worth acting on, and smaller than the headline implies.

The verdict is that this changes an account, not a budget. If your category expertise is published under a logo, it’s sitting in the least-cited social account type, and moving it costs nothing but a byline decision and the willingness to let named employees speak for the company. Pick the three people who actually know something, put their names on it, and check in 90 days whether anything of yours gets quoted. That test costs a quarter and settles the question for your company rather than for Oktopost’s sample.