The Maze: A shopping assistant can compare products without changing where its owner wants to buy. In an August 2026 PYMNTS Intelligence survey of 2,191 US consumers, shoppers imagined an autonomous AI assistant buying the same item at identical prices, shipping and return terms. A large marketplace was the preferred destination for 40.2%; a brand's website attracted 16.1%. Giving software the shopping task does not automatically erase the customer's seller preference.
Marketplaces lead even when the offer is equal. The destination exercise held the product and commercial terms constant. Large online marketplaces led at 40.2%, followed by department stores or large retailers at 21.8%. Another 21.8% would let the assistant choose, while 16.1% selected the brand's own website. These are shares of respondents expressing preferences, not market shares of purchases made by agents. The marketplace lead over brand sites is 24.1 percentage points. A familiar place to transact can remain attractive even when an assistant does the searching.
The undecided group leaves room for competition. The 21.8% willing to let AI choose equals the share choosing a large retailer and exceeds the share choosing a brand site. That is a potential opening for competing merchants, not proof that agents will award them orders. The survey does not establish how an assistant would rank sellers, which services it could access or how its owner would approve payment. Nor is 40.2% a majority: the remaining respondents chose another destination or left the decision open. Brands have a disadvantage in this exercise, not a verdict of extinction.
Shopping advice and checkout have different competitive strengths. In a separate question about consumers' most recent AI-assisted purchase, 59.7% completed it on Amazon and 6.3% on a brand's website or app. Here, AI-assisted means AI helped with shopping; it does not mean an agent independently placed the order. This different group and question cannot be compared directly with the hypothetical destination percentages. It does, however, illustrate the same commercial tension: a merchant can retain the transaction even when another service helps the customer choose. Losing a product-search interaction need not mean losing the sale.
A better direct offer matters more than automatic checkout. Among marketplace shoppers, 53.9% said lower prices or exclusive discounts could move them to a brand or store directly; 49.9% cited free shipping as fast as the marketplace. AI-handled checkout appealed to 12.1%. These are separate incentives, not slices to add together. The survey measures what people say would persuade them, not the conversion lift from implementing each feature. Still, it challenges a roadmap that treats automated payment as a substitute for competitive delivery and pricing. Software cannot supply a discount or faster parcel on its own.
Why it matters: The strategic split is between helping a shopper choose and earning permission to complete the purchase. If these US preferences persist in actual agent use, marketplaces could keep transaction relationships while AI services take more of product discovery. Brands would then need both visibility in recommendations and a compelling reason to buy direct. That could require discounts or faster shipping, which cost money; adding an assistant alone does not remove that tradeoff. Other markets may differ, and a hypothetical survey cannot predict which agent or merchant will win.
