How AI Search Chooses Which Stores to Cite

In two steps. First comes retrieval: the engine runs web searches for the shopping question and pulls back pages that look relevant. Google's AI features guide says AI Overviews and AI Mode may use "query fan-out", which means one question is split into several related searches, such as price, reviews and where to buy. Then comes generation: the model reads what it found, writes the answer and chooses which sources to name and link.

To cite means to name a source and link to it in the answer. OpenAI's web search guide says the model decides whether to search at all, and that the sources it consulted are often more than the citations it shows. So being found is not the same as being cited. A store first has to be in the pool of pages the engine reads, then be the clearest, best-backed source in that pool.

The two steps are the same; the evidence is different. A B2B buyer asks for a supplier, and the engine leans on directories, case studies and company facts, as our guide on how AI chooses suppliers explains. A shopper asks for a product, and the engine needs product facts it can compare: price, stock, specs, shipping, returns and what other buyers say.

A shopping answer also often names a product first and a store second. The same item may be sold by the maker, a marketplace and many shops. When the facts match across those places, the engine can pick any of them as the link. When only one store states them clearly, that store has a better chance of being the one that is cited.

It needs to be reachable by each engine's crawler, the program that fetches pages for search. OpenAI's crawler documentation says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers. A robots.txt rule or a CDN bot filter that blocks it by mistake removes the store from the pool. Our guide to what each OpenAI bot does shows the exact lines to allow.

For Google, the bar is ordinary search. Google's AI features guide says a page must be indexed and eligible to show with a snippet to appear as a supporting link, with no extra technical requirements. The same page adds that meeting every requirement does not mean Google will crawl, index or show the content. Eligibility gets a store into the pool. It does not decide the choice.

Neither OpenAI nor Google publishes a ranking formula for stores in AI answers. What they do document, plus what we see in our own tracker rounds, points to a few signals a store controls. The table below lists them, why each one matters and how to check it. Treat it as a checklist, not a ranking model.

Because the model writes its answer from text. A product page that states what the item is, who it is for, what it costs, and how fast it ships gives the model sentences it can repeat. A page that hides the specs in an image or a script gives it nothing. Our guide to getting products recommended by ChatGPT shows the page-by-page method.

Product markup (structured data that labels name, brand, price and stock for machines) helps a machine read the same facts without guessing. Google's AI features guide says no special markup is needed for its AI features but asks that markup match the visible text. Our guide to product schema for AI shopping answers covers the fields.

Because those sites already gather what a shopper asks about: many sellers, many prices and many reviews on one page. An engine that wants proof for "best rated" or "most reliable" finds it there first. A store competes for the link by stating its own facts clearly and by being described accurately on those same sites. In our tracker data, a marketplace domain often sits next to the store domain in the list of cited sources.

Reviews matter to people as well. The Medill Spiegel Research Center found that a product with five reviews was 270% more likely to be purchased than one with none (study first published 2017). That is a finding about shoppers, not about AI engines. It still points the same way: collect real reviews from real buyers, and never buy or write your own.

Not on its own, and most stores cannot send one yet. A product feed is a file that lists every product with its price and stock. OpenAI's commerce guide says feeds give ChatGPT the catalog data it needs to index products and present accurate product information in shopping experiences. It also says feed onboarding is open to approved partners only.

So for most stores the open web is the main way in today: crawlable pages, plain facts and mentions on other sites. If you already keep a clean feed for Google Merchant Center, keep it accurate; it is the same product data an AI feed would need later. Running WooCommerce? Our guide to GEO for WooCommerce stores shows what the platform already prints and what to add.

On 2026-10-25 we asked US buyer questions about 7 Netbase JSC brands on three engines. Questions that contained the brand name named it in 62 of 84 runs (73.8%, 95% Wilson interval 63.5-82.0%). Questions without the brand name, the ones a new shopper asks, named a Netbase brand in only 20 of 274 runs (7.3%, 4.8-11.0%). A Wilson interval is the range the true rate most likely sits in.

CMSmart (cmsmart.net, part of Netbase JSC) sells WooCommerce plugins online. Its domain was the most-cited in its niche in the US, with 96 citations against 89 for woocommerce.com, the marketplace. In the UK it was named in only 2 of 22 runs. One brand in one round shows that a store can out-cite a marketplace in one market and vanish in another; it does not show why. The CMSmart case study has the full figures.

When you want to know which questions name you, which sources the engines cite instead of you, and whether a fix moved anything. One answer in your own ChatGPT account cannot tell you that, because answers change from run to run. Leadee is an AI-native platform that stands on the shoulders of giants: it asks the engines themselves through the official APIs of OpenAI, Google and Perplexity, and reads your pages the way their crawlers do.

How Leadee runs the full store audit as a service is on our AI visibility for ecommerce stores page. To pick a tool that tracks shopping questions over time, see our overview of AI visibility tools, which covers Leadee and other market players side by side. Our methodology page explains how runs and intervals are counted.