Here's the answer: in the contested answers where purchases happen, AI reaches for third-party evidence, and either finds yours or finds your competitor's. The research is blunt about it. Muck Rack's 25-million-link analysis puts earned media at 84 percent of AI citations in its pool. Ahrefs finds top-quartile mention volume worth roughly 10 times the AI visibility. And the Princeton study that named the GEO field measured that content carrying quotations, statistics, and citations gains up to 40 percent more generative visibility.
So build the earned layer like architecture, targeted where the machine already looks. Make every press placement citable: pitch with a proprietary number or a crisp definition attached, because journalists quote data and models re-quote journalists. Get named in the headline or first paragraph, where extraction happens. Then bank every placement on a press page so the earned and owned layers reinforce each other.
Reviews stopped being social proof and became retrieval data: ChatGPT now surfaces Yelp reviews, ratings, and photos directly through the OpenAI and Yelp partnership. Ask consistently, respond to everything, and concentrate on the platforms your category's answers actually cite.
And if your founder outscores the brand, wire the transfer: every founder placement names the brand in the same breath, and the site connects the two on the record. Authority that stays personal is rented. Transferred, it compounds.
One warning: evidence architecture builds on things that are true. Manufactured awards read as exactly what they are.
The full earned-layer manual is Volume 4 of our Primacy Playbook.
The machine has something to cite now.