ChatGPT is becoming a storefront and most brands are absent

OpenAI has turned ChatGPT into a place people shop, not just ask questions. Through its shopping-research feature, in-chat checkout, and a merchant program built on the Agentic Commerce Protocol, ChatGPT now recommends products and lets users buy inside the conversation. The problem for brands: when the answer is a shortlist of one or two names, most businesses are not in it. An analysis of nearly 190,000 ChatGPT results for restaurant queries found roughly 83% of restaurants absent, against about 14% on Google.

What happened

OpenAI has added a set of commerce features to ChatGPT over 2026. Shopping research produces a personalised buyer’s guide from a plain-language description of what someone wants. Instant Checkout, the “Buy it in ChatGPT” flow, lets users complete a purchase without leaving the chat, using the Agentic Commerce Protocol so that merchants can share product feeds and promotions directly. At launch Instant Checkout covers US Etsy sellers, with Shopify merchants expected to follow.

The visibility gap is not explained by brands being offline. The same restaurants that vanish from ChatGPT largely maintain a presence on Google. They have done the traditional local-search work and still disappear the moment the question is routed through an AI model instead of a results page.

Why it matters

A search results page shows ten links and lets the user choose. An AI assistant collapses that into a recommendation, often a single name, and the brands it names capture a disproportionate share of the attention. There is no page two to rank on. Either the model has clean, structured, current data about a business and its products, or that business does not exist in the answer.

That shifts the job. Being findable in AI assistants depends less on classic ranking signals and more on machine-readable accuracy: consistent location data, complete attributes, and, for commerce, a structured product feed the model can trust.

What this means for multi-location retail brands

For a retail chain, this is a data-quality problem at estate scale, not a content-writing one. If opening hours, addresses, categories, and product availability differ between a brand’s own site, its local business listings, and its feeds, an AI model has no reliable version to surface, so it defaults to the competitors whose data is coherent.

The work is to make one accurate, structured version of every location and product available everywhere a model might read it, then keep it current as stores, ranges, and hours change. That is the same discipline behind being cited in AI search results, and it is what a system like Places AI is built to manage across hundreds or thousands of locations rather than one at a time. It builds directly on the pattern in our earlier report that ChatGPT recommendations drive real site visits: the traffic is real, but only for the brands the model can actually see.

In an analysis of nearly 190,000 ChatGPT results for restaurant queries, about 83% of restaurants did not appear at all, compared with roughly 14% on Google.

Local Falcon, AI visibility study, 2026

The bottom line

ChatGPT is turning into a discovery and checkout channel, and the brands it recommends are the ones whose data it can read cleanly. For multi-location retailers, the priority is not writing more content, it is making sure every location and product exists in accurate, structured form before the model decides who to name.