Why earning mentions matters more than polishing your own pages
The structure and markup were refreshed for current answer engines; the original analysis is preserved as written.
It is tempting to think AI visibility is won on your own website. The evidence from 2024 says otherwise: when AI search answers a question about your category, it overwhelmingly cites third parties — press, reference sites, communities — and rarely your own pages. Which makes earned media, not just on-site polish, the lever.
the short answer
AI search cites third-party sources — press, reference sites, communities, review platforms — far more than a brand’s own pages, because a model treats independent corroboration as more credible than self-description. So appearing in AI depends heavily on earned media: getting mentioned, accurately and consistently, by the sources the model already trusts. Digital PR is the work of earning those mentions. Your own site still matters as the anchor of your entity, but it is a claim, not the outside proof.
key takeaways
- Audits in 2024 (such as Li & Sinnamon) find AI search leans on third-party sources — press, reference sites, communities — and rarely on a brand’s own pages.
- So appearing in AI depends heavily on earned media: being mentioned by sources the model already knows and trusts, which it reads as independent corroboration.
- Digital PR for generative engines is the set of plays that earn those mentions: original data, expert commentary in the press, presence in third-party reviews and comparisons, consistent profiles on reference sites.
- Your own site still matters — it anchors your entity with a clear, direct account — but it is a claim, not the outside confirmation a model weighs more heavily.
- Consistency across sources strengthens the model’s confidence; contradictory descriptions weaken it and make it hold back.
where the answer’s citations come from
Read the picture and the strategy follows. The thick lines — press, reference sites, communities, review and comparison platforms — are where the answer’s citations actually come from. Your own site is the thin line: essential as the anchor that defines who you are, but rarely the source the answer names. Visibility, then, is mostly earned out there, not built in here.
Why this reorders the budget
Most teams spend the majority of their content effort on their own site, because that is the surface they control. That instinct made sense when the goal was to rank that site in search. It makes less sense when the goal is to be cited in an AI answer, because the answer is mostly assembling its sources from everywhere except your site. A brand can have an immaculate set of pages and still be nearly absent from AI answers if no credible third party is talking about it — and a competitor with a plainer site can be everywhere the model looks because the press, the reference entries, and the review sites all describe it. The uncomfortable implication is that some of the effort poured into perfecting owned pages would do more good earning mentions off them.
This is not an argument to abandon your site; it is an argument to rebalance. The site still has to be the clear, current anchor that earned mentions point back to, and that work is real. But once the anchor is solid, the marginal hour is often better spent on the things that put your name, accurately, into the sources a model already trusts — original data worth covering, expert voices worth quoting, a presence worth listing on the sites buyers consult. For a business that has historically thought of content as a website project, that is a genuine shift: visibility in AI is earned at least as much in other people’s publications as on your own.
Earning the mention, in three parts
Why the model leans on third parties, the concrete plays that earn those mentions, and where your own site still fits. Open each layer for the part that changes how you spend effort.
01 Why AI leans on third parties
A generative model is, in part, a consensus machine. When it answers a question about your category, it is drawing on what many sources have said, and it is more willing to repeat a claim that several independent, credible sources agree on than one that appears only on the page of the company it concerns. That is a reasonable instinct — corroboration is how anyone sorts reliable information from marketing — and it shows up in what these systems cite. Audits through 2024 found that AI search tools leaned heavily on commercial and news sources, on established reference material, and on community discussion, while a brand’s own pages were rarely the thing being cited. The lesson is uncomfortable but clarifying: the model does not take your word for who you are. It assembles its picture of you from what the rest of the web says, which means your visibility is, to a large degree, in the hands of sources you do not own.
02 The digital PR repertoire for GEO
If the mentions live on third-party sources, the work is to earn them — and the plays are concrete. Original research is the strongest: publish real data the model and the press both want to cite, and you become a primary source others reference, which we have written about separately as a citation magnet. Beyond that, expert commentary places your people in the press as named authorities on your topic; presence in the category’s review and comparison sites puts you where considered buyers and the models serving them go to compare; and accurate, consistent profiles on reference and directory sites give the model a stable description to anchor on. The throughline is substance: each of these earns a mention because there is something genuinely worth mentioning, not because of a placement trick. Done together, they surround your name with credible, consistent, independent descriptions — exactly the raw material a model uses to decide it knows you well enough to bring you up.
03 Where your own site still fits
None of this demotes your own site; it reframes its job. Your pages are the anchor of your entity — the place where you state clearly what you do, answer the genuine questions in your field, and give a current, unambiguous account that everything else can point back to. When a third-party mention sends a model looking for who you are, your site is what it should find: coherent, authoritative, and consistent with what the outside sources say. The failure mode is treating the site as the whole strategy, pouring effort into pages no model cites while ignoring the earned mentions that actually carry weight. The complementary failure is earning mentions that point back to a vague or contradictory site, so the corroboration has nothing solid to confirm. The strong position holds both: a clear site as the anchor, and a web of accurate third-party mentions as the evidence — the same balance of self-description and outside proof that the AC Group has helped clients build for {years} years.
Where to start: audit the footprint you already have
Before earning new mentions, it is worth seeing the ones you already have, because the picture is usually lopsided in instructive ways. Ask a few of the AI tools the questions a buyer in your category would ask, and watch which sources the answers lean on and whether your name appears at all. Most brands discover one of two things. Either they are largely absent — the model builds its answer from sources that do not mention them — or they are present but described inconsistently, named one way on a reference site, another in the press, a third on a review platform, so the model has no stable account to settle on. Both are fixable, and seeing which one you have tells you where to spend.
If the problem is absence, the work is earning the first credible mentions in the source types that matter for your category — usually starting with original data worth covering and presence on the review and reference sites buyers consult. If the problem is inconsistency, the faster win is to make the existing mentions agree: correct the reference entries, align how profiles describe you, and give the press one consistent account. The order matters because effort spent earning new mentions is wasted if they will only add to a contradictory picture. Get the description consistent first, then widen it — that sequence turns a scattered footprint into the coherent entity a model will repeat.
Digital PR for AI: quick answers
Isn’t this just old-school PR?
The goal is older than AI — earn coverage and mentions from credible third parties — but the reason and the target are new, which changes how you do it. Classic PR chased a placement for the human readership of the outlet and, later, for the backlink. Digital PR for generative engines chases the same placement because the model reads those outlets too, and treats a mention in one as a signal about you. That shifts the emphasis in practical ways. You care less about the single splashy hit and more about consistent presence across the sources a model is likely to draw on; you care that the mention describes you accurately and consistently, because the model is learning what you are from the words around your name; and you value reference and community sources that traditional PR often ignored. So the instinct is familiar, but the aim is to shape what the model understands and is willing to repeat about you, not only to reach the outlet’s audience directly.
Does my own content still matter?
Very much — it just cannot do the whole job alone, and pretending it can is the mistake. Your own pages are where you state plainly what you do, answer the real questions in your category, and give the model a clean, current, unambiguous account of your entity; without that, third-party mentions have nothing coherent to point back to. But your site is, by definition, you talking about yourself, and a model weighs that differently from independent corroboration. When several outside sources describe you consistently, the model reads that as consensus and grows confident enough to repeat it; your own page alone is a claim, not a confirmation. So the right model is both-and: a clear, authoritative site as the anchor of your entity, and earned mentions as the outside evidence that the anchor is real. Neglect either and you are weaker than a competitor who does both.
Which third-party sources matter most?
It depends on the engine and the question, and in 2024 the honest position is that the precise rankings were still being mapped — so the wise move is breadth across types rather than betting on one. Broadly, the sources that recur are reference sites that act as neutral descriptions of entities, established press and trade publications that carry editorial credibility, community and discussion platforms where real users compare options, and the review and comparison sites specific to your category. Different engines lean on these in different proportions, and those proportions were shifting as the tools matured. Rather than chase whichever source is most cited this month, the durable approach is to be accurately and consistently present across all of these types, so that whichever sources a given engine reaches for, it finds a coherent account of you. For a B2B audience the category-specific review and comparison sites tend to carry unusual weight, because that is where considered buyers and the models serving them go to compare.
How is this different from link building?
Link building optimised for a machine that counted links as votes; digital PR for AI optimises for a machine that reads sentences as meaning. The old game could be satisfied by a link with little regard for context — the link itself carried the value. A generative model does not need a hyperlink at all to register that you exist and what you are; an unlinked mention in a credible source still teaches it something, and a mention that describes you well is worth more than a bare link that says nothing. So the craft moves from acquiring links to earning accurate, substantive mentions in trusted places — coverage that actually says who you are, what you do, and why you are credible. The overlap with old link building is real, but the unit of value has changed from the link to the described mention, which rewards genuine newsworthiness and substance over the mechanics of placement.
A note on sources and certainty
This is written in October 2024. The claim that AI search leans on third-party sources rests on the audits available then — work like Li and Sinnamon’s 2024 study of ChatGPT, Bing Chat and Perplexity, which found citations dominated by commercial and news sites rather than academic or owned pages. We have deliberately not quoted the precise source-share figures that circulated later — which domain accounts for what percentage of which engine’s citations — because those large-scale measurements were assembled afterward, and borrowing them backward would lend a false precision to an October 2024 view. What was already clear, and is what this note rests on, is the direction: models corroborate across independent sources and rarely cite a brand’s own pages, so earned media is central to AI visibility. That has been the AC Group’s working assumption for 27 years — that authority is granted by others, not asserted by yourself.