Representation forms where you don't publish

Only 1% of the sources cited by LLMs comes from brand-owned sites. The other 99% of the representation gets built on material you do not control.

The figure is stark and belongs up front: only 1% of the sources cited by LLMs comes from brand-owned sites (McKinsey, 2026). The rest is earned — forums, reviews, comparisons, third-party articles.

The order of magnitude is not new: even before AI, reputation was largely built outside owned channels. What changes is the synthesis mechanism. Before, earned sources were one of the things a user might encounter, in a list where the brand’s own site appeared anyway, usually near the top. Now they are read, weighted and summarised by a system that returns one answer, and in that answer owned material weighs what the figure says: one hundredth.

The corpus already records the operational consequence: digital PR “moves from a support activity to the infrastructure of AI visibility”. This is not a departmental promotion, it is a shift in where the lever gets applied. Work on the site does not stop mattering — it remains the source the others cite when they need to verify something — but it stops being the place where representation forms.

There is a second level, less visible and more uncomfortable: a substantial part of what forms the interpretation is not even observable. The corpus says so in the Social Listening entry of Trust, AI literacy and calibration: “the most influential conversation about the brand happens in private chats, unobservable” — the user compares alternatives, complains, chooses, and nobody is listening. It is “a structural loss of visibility, not one better tooling closes”. So the 99% is not only material you do not control: it is partly material you do not see.

The decision

The question. How much investment do we move from what we publish to what we do not control?

The default by omission. Nothing moves, because owned is the only line item with an owner, a calendar and a number. The budget stays where a dashboard exists, and the lever worth 99% has neither owner nor measure. Nobody decides to neglect earned: you simply carry on funding what you know how to report on.

The cost of either branch. Moving investment to earned means paying for something you do not control but do influence: reviews, presence in industry forums, authoritative mentions, consistency of brand data across sources. None of these produces an attributable result, and all of them produce an attributable cost — the classic shape of the spend that loses the argument in a meeting. Not moving it means carrying on tending the one place where representation does not form, and accepting that the 99% gets written by other people. With one aggravating factor: if the representation is wrong, the corpus records that “a hallucination about the brand is reputational damage with no channel for correction”, and the only available lever is correcting the sources upstream — which is exactly the lever that was not funded.

How it is measured. Share of presence is measurable, with the caveats Staying in AI answers lists: a prompt set built by intent, versioned, with enough repetitions. Position is not. And the part of the conversation that happens in private is not measurable at all, by any tool present or announced. On this decision you accept working with a known partial coverage, not with complete data.

Who decides. Digital PR / Reputation for the lever, but the budget is not theirs. It is a reallocation across line items belonging to different functions, and it has to be decided where channel budgets are allocated — with the AI Visibility Analyst supplying the measure, where the figure exists.

→ a particular case of the INTERPRET position, the one with the least control of the four.

Effect on the professions

Digital PR / Reputation — The role moves from a support activity to the infrastructure of AI visibility. Reviews, presence in industry forums, authoritative mentions and consistency of brand data across sources become primary GEO levers, not peripheral activities. The change of mandate is bigger than the change of activity: the same things you did before are now the main channel, and have to be funded as such.
Social Listening / Brand Intelligence — The object of listening moves upstream: the sources that feed generative answers become more relevant than traditional social channels. Two caveats the corpus records and that are better budgeted for: the quality of the social signal drops because it is diluted by synthetic content, so mention volume loses value as an indicator; and the most influential part of the conversation, the one in private chat, stays beyond the reach of any tool.
Content Strategist — Owned content changes function: it is no longer where representation is built, it is the source the others cite when they need to verify something. That makes it more important, not less, but for a different reason than the one it used to be written for: verifiable accuracy and consistency with what is found elsewhere matter more than persuasion.
SEO / GEO Specialist — The perceived authority of third-party sources becomes a visibility factor on a par with the semantic structure of your own pages. It is the point where GEO and digital PR stop being separate disciplines, and neither covers the lever on its own.
AI Visibility Analyst — This is the figure that should supply the measure this decision is made on, and the corpus declares its instrumentation unreliable for the part most in demand: position. What it can give is share of presence and its variation over time, with the caveat that 40–60% of cited domains change within a month even when nobody does anything.