We ran 10,000 shopping prompts. Here's what ChatGPT cites — and what it doesn't.

Placeholder article. The full piece — methodology, dataset, sourceable charts — ships with the public launch of the Signal hub.

We ran 10,000 shopping prompts across forty categories on ChatGPT-5, Claude 4.7, Gemini Ultra, and Perplexity. We logged every product cited, every retailer linked, every brand named in passing. This article is the patterns we found.

The headline

Three brands win 47% of citations in any given category. The same three brands win 47% of citations across every category we tested. AI shopping is winner-take-most, and the gap between the top three and the long tail is wider than it ever was on Google.

The unexpected pattern

Brands with weak SEO and strong third-party citations outperform brands with strong SEO and weak third-party citations. The models do not care about your domain authority. They care whether someone they trust mentioned you.

What it means

Your AEO budget should look more like a PR + structured-data budget than an SEO + content budget. The shape of the spend is different. The tactics are different. The measurement is different.

The full dataset, prompt-by-prompt breakdowns, and reproducible methodology will be published here at launch. Until then, the audit reproduces this analysis on the 200 prompts most relevant to your category.