AINEWS
Holiday 2026 will be the first mass test of AI that shops on your behalf. Survey data shows shoppers are drawing a clear line at the buy button, and that line decides who keeps the cart.
The shopper asks an AI assistant to shortlist a $300 jacket across three retailers, gets a clean comparison in a chat window, then opens the merchant's own site to pick the seller and complete the purchase. The discovery happened inside the AI. The decision and the payment happened inside the merchant. That two-step is the 2026 pattern 132 million U.S. adults are already running, and holiday 2026 will be the first mass test of who owns the cart at the end of it.
PYMNTS Intelligence data, drawn from the firm's consumer survey and a closed-door roundtable hosted at Mastercard's New York technology hub, captures the shape of the pattern. 58.4 million U.S. adults now begin online retail research with AI, up 18% since June 2026. The discovery surface has gone conversational. But only 24% of respondents would let an AI agent shop and buy for them. The buy button is the line.
That gap is the structural number to watch this season. It is not a trust deficit and it is not a hype curve. It is a delegation boundary: shoppers are choosing which steps to hand off and which to keep, and that choice is consistent enough across the panel to call a category shift rather than a curiosity.
Three things sit on the merchant side of that line.
The first is ownership of the relationship. 59% of AI-assisted purchases still end at Amazon, per the same PYMNTS panel. When discovery happens in an AI tool but checkout happens on a retailer's owned surface, the retailer is buying the customer each time. The AI is a top-of-funnel. Whoever runs the post-purchase experience, including returns, refunds, and support tickets, owns the next season's cart.
The second is inventory plumbing. Among 60 U.S. merchants the firm surveyed, 53% said they are prepared to share real-time inventory data with AI systems. Only 23% said they are prepared to connect directly with AI platforms or agent networks. The asymmetry is the whole story: retailers can show stock, but most still cannot transact through the conversational surface where shoppers are starting their search.
The third is the card. 52.3% of consumers in the PYMNTS sample prefer to keep their payment card outside the AI app. 14% would let the app store it. The middle is the design problem. A conversational checkout that takes a saved card on file, routes the authorization through the merchant's existing rail, and returns a clean confirmation to the chat window is the product retailers and payments companies have to ship before next season. One that holds the card in the AI app and tries to act on the shopper's behalf is a product most shoppers have not agreed to.
Per PYMNTS' writeup of the Mastercard roundtable, 40 senior leaders from retail, payments, financial services, technology, and AI converged on the same boundary from the sell side. The 18% jump in AI-led research in three months was treated as evidence the discovery surface has moved. The 24% willing to delegate the buy was treated as the constraint to design around, not a number to argue with. A customer who keeps the final click is a customer to earn each time, not a customer to convert.
Two numbers to watch when post-holiday panels land. Whether the 24% who would let an agent close the purchase moves up. That would mark a structural shift, not a survey artifact. Whether the 23% of merchants prepared to connect with AI platforms crosses 50%. That would mark the build-out catching up with the search behavior. Either move changes the next season. Neither is a given.