On October 2, Lightspeed Commerce published an analysis of 460,000 AI shopping responses across ChatGPT and Google's AI products. EMVCo closed public comments on its Agentic Payments Framework for Specifications on September 30. Read together, they show who gets named when a shopper asks an AI assistant what to buy, and whether the payment clears.
Vaer AI and Lightspeed Commerce found that bigger merchants have an edge from the very start of AI search. When ChatGPT included retailer names in its own web searches, 97% were large or mid-sized chains. Small and local retailers accounted for just 1 to 4%. That pattern holds even when the assistant can see smaller stores. Large and small retailers each made up roughly 38% of the sources AI cited, but big chains received 52% of the top recommendations. Smaller businesses got 20%.
One example from the study makes the point. A shopper asking for office supplies in New York saw three small local retailers among the six sources ChatGPT cited. The first recommendation was Staples, with the local options listed below it.
The pattern repeats across assistants. Alexa for Shopping showed toys from 177 brands, and nearly every one came from Amazon's own store. The same toy request to ChatGPT produced recommendations from 24 retailers. Google's AI Mode went to 37.
When the assistant's default answer draws from a narrow set of large merchants, a brand outside that set has to earn its place in the answer before its product pages matter. The EMVCo work addresses the second half: what happens after the assistant names a product. EMVCo's proposed Intent Services would let payment participants register, reference, retrieve, and manage consumer-authorized intent before, during, and after a transaction. The standards body may also develop know your agent capabilities and transaction indicators to identify agents and signal their involvement in payments. A shopper could tell an AI agent to buy a specific product, spend no more than a certain amount, or make recurring purchases. The framework would carry that authorization through the card payment.
PYMNTS Intelligence found that 56% of consumers would let AI agents search and compare products, but just 37% would let them authorize payments. Search and comparison are already a consumer behavior. Payment authorization is the next step, and the standards are being written now.
For a marketing or ecommerce leader, the practical question is whether your brand can be seen and whether it can complete the sale. Ask the assistants the questions your buyers ask and note which retailers get named. Then check whether your payment flow can accept an AI agent's transaction.
If the answer is a national chain or the assistant's default retail partner, the shelf you need to be on is the store the assistant already looks at. The EMVCo framework is moving from experiment to infrastructure. The brands that verify this now will be the ones whose authorization survives when the standards go live.
Common questions
What does the Lightspeed study actually show?
Lightspeed Commerce and Vaer AI analyzed 460,000 AI shopping responses across ChatGPT and Google's AI products. They found that larger merchants get recommended more often, even when smaller retailers are present in the sources the assistant cites.
Why does the EMVCo framework matter to a brand?
EMVCo's proposed Intent Services would standardize how consumer-authorized intent moves through card payments. If an AI agent makes a purchase on a shopper's behalf, the framework defines how that authorization is registered, referenced, and managed.
What should I do first?
Ask the AI assistants the questions your buyers ask and note which retailers get named. If your brand is not among them, the next step is to be stocked where the assistant already looks.
Is this a search problem or a payments problem?
It is both. First, the assistant has to see and cite your product. Second, the buyer's authorization has to survive the transaction. The two developments address different halves of the same distribution challenge.