The most striking number in Riskified’s latest consumer survey is not the one about adoption. It is the one about refusal.
Sixty-one percent of consumers now use AI tools for product discovery and recommendations, according to the fraud prevention company’s Q1 2026 Agentic Commerce Pulse survey of 2,000 shoppers in the United States and United Kingdom. That figure tracks with the broad direction of every major retail platform investing in AI-powered search, chat-based recommendations, and automated product comparison.
But 55 percent of those same consumers say they are not comfortable letting an AI agent complete a purchase on their behalf. And 46.5 percent do not trust any company — AI platform, retailer, or otherwise — to manage purchases for them at all.
That gap between discovery comfort and transaction refusal is where the agentic commerce story currently lives. And it is widening.
A quarter earlier, Riskified’s Q4 2025 survey of 5,400 consumers across multiple global markets found 70 percent expressing at least some comfort with AI agents making purchases. The drop to 55 percent discomfort in Q1 2026 should be read with a caveat: the two surveys use different sample sizes and geographies, making direct comparison approximate. But the directional shift is hard to ignore, especially when paired with the finding that 53.9 percent of respondents now believe AI could increase the risk of online fraud.
As Jeff Otto, Riskified’s chief marketing officer, put it: “Consumers are clearly embracing AI as a shopping assistant, but they’re drawing a firm line when it comes to autonomy and accountability.”
The accountability question is where the data gets structurally interesting. When asked who should be responsible when an AI agent makes an unauthorized or erroneous purchase, 50.8 percent of consumers pointed to the AI platform. Only 23.2 percent said the retailer or brand. Just 18.7 percent were willing to accept personal responsibility.
That expectation — that the platform should guarantee the agent’s behavior — has no current home in payment infrastructure or regulation. The GENIUS Act, the closest thing to a federal framework for digital asset transactions, regulates stablecoin issuers. It says nothing about who bears liability when an AI agent initiates a transaction that the consumer did not intend. The gap between consumer expectation and legal reality is, for now, a vacuum.
Product.ai’s April 2026 Trust in AI Commerce Report, which surveyed 1,463 U.S. online shoppers independently, reinforces the pattern. Only 14 percent of consumers who have used AI for product research trust AI recommendations enough to execute a purchase without verification. Eighty-six percent check before buying. Forty-two percent will not trust AI for purchases over $25.
The convergence of these two datasets — one from a fraud prevention vendor, the other from an independent research firm — points to the same structural ceiling. AI is welcome in the browsing aisle. It is not welcome at the register.
Riskified has a commercial interest in making this case. The company sells chargeback guarantees to merchants, and framing agent-driven commerce as a fraud vector serves that business. But the underlying consumer sentiment data, corroborated by Product.ai’s independent findings, describes a real market constraint rather than a manufactured one.
The practical signal for merchants: 73.9 percent of consumers now expect biometric verification or one-time passwords for every transaction an AI agent attempts. The path to adoption runs through more authentication friction, not less. Agentic commerce, at least as consumers currently understand it, is a discovery tool that ends with a human clicking buy.