How AI chooses which brands to cite and how to become one of them

The numbers have stopped being debatable. According to McKinsey's October 2025 study, 44% of consumers now name AI search as their primary source of information before buying, ahead of traditional search at 31%. Around half of Google searches already display an AI-generated overview, a proportion expected to reach 75% by 2028. And Previsible, which analysed nearly two million AI-driven sessions in 2025, measured a 527% growth in AI-referred traffic in five months. The question is no longer whether your customers are asking AI — it's what AI answers about you.

This is the uncomfortable discovery most brands make: according to McKinsey, a brand's own website accounts for only 5 to 10% of the sources AI cites in a given category. PR firm 5W puts it even more starkly: 85.5% of AI citations come from earned media — press, creators, communities, reviews — not from brand websites. The logic is structural: an AI is designed to synthesise credible, independent perspectives, not to amplify advertising copy. A brand calling itself excellent weighs less than three independent voices saying it instead.

Each engine has its own mechanics — Google's AI overviews build on the Google index, ChatGPT on Bing's, Perplexity on live web retrieval — but all apply the same selection logic: expert content, signed by identifiable authors, published on trusted third-party domains, and corroborated across sources. That last point is decisive: an AI is far more likely to cite a claim when it finds it, independently phrased, across several authoritative sources. One article is a signal; five concurring voices on five platforms form an answer.

If 85% of citations come from earned media, then the craft of getting credible third parties to talk about you — journalists, creators, experts — has mechanically become the central lever of AI visibility. A press article that recommends, a creator's long-form video that demonstrates, a transcribed podcast appearance, an authentic community conversation: these are exactly the formats AI indexes, weighs and cites. That is the principle behind our Matr·IA·chka offer: activating the five source families AI consults, and making the signals converge rather than producing just one.

Classic SEO indicators no longer suffice: most AI users never click, so the influence happens upstream of the visit. The reference metric is share of answers — the percentage of AI responses, across a basket of questions representative of your category, that mention your brand. To that you add citation source analysis (which URLs AI uses when it talks about you), competitive benchmarking, and the growth of branded searches: when an AI cites you, users often check your name on Google right after.

Two acronyms for the same discipline: GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization) refer to the set of practices maximising a brand's presence in generative AI answers — ChatGPT, Gemini, Perplexity, Google's AI overviews. Where SEO optimises for a ranking of links, GEO optimises for a citation inside an answer.

No — it has become a necessary but insufficient foundation. AI systems rely on Google's and Bing's indexes: content invisible to search engines is invisible to AI too. But ranking is no longer enough: you need to add the layer of third-party credibility AI requires before citing a brand.

Five main families: authority media, social networks and their creators, podcasts, communities and forums, and knowledge bases such as Wikipedia and Wikidata. Each family weighs differently depending on the engine and the category — hence the value of working them together rather than betting on just one.

AI visibility is built through accumulated signals, not one big hit. First citations can appear within weeks when the brand starts from a solid base; establishing a durable presence in a competitive category is managed over several months, with regular measurement to redirect the effort.

Yes — that is one of the good news stories of this shift. AI favours relevance and corroboration over raw fame: a precise brand, well documented by credible third parties in its niche, can get cited where a bigger generalist isn't. The first-mover window is still open.

Last updated: September 2026