Can you trust an AI to shop for you? Consumer Reports on building “loyal” agents

AI agents can now check out for you — inside ChatGPT, and through new payment rails from Visa and Mastercard. The technology is here; the trust is not. Across 2025–2026 surveys, only a minority are comfortable letting an agent actually buy: Bain found just 24% of US consumers comfortable completing a purchase through AI, and only 10% have done it. A Riskified survey in early 2026 put 55% as still not comfortable handing over the buy.

A Consumer Reports and Stanford workshop — the second in its “Loyal Agents” series — asks the question underneath that hesitation: will these agents genuinely serve you, or become “sophisticated sales funnels that prioritize profit over user needs”? It names what has to be solved first — secure payments, who is responsible when an agent makes a mistake, and whether the agent actually reflects your preferences — and floats a striking idea: that a shopping agent should owe you real fiduciary and legal duties, the way a professional does.

The trust-builders are consistent across the research. Transparency into how the agent decides — Visa found nearly 90% of people want it. A human kept in the loop, with clear spending limits and an explicit approval step. Control over your data. And, hardest of all, demonstrable neutrality — so the agent isn’t quietly steering you toward whoever paid for placement.

That is the bar OwnWise is built to meet. OwnWise is an AI, but a transparent one: it shows its reasoning, its sources, and its uncertainty, and it is merchant-neutral — it holds no inventory and has no product to push (and if it ever earns referral compensation, it discloses it and doesn’t let it drive the recommendation). The goal isn’t to hand off the decision — it’s to help you make it, and to show its work so you can check it. See also whose side is your shopping agent on? and the new rails that let an agent check out.