Key Takeaways:
- 72% of U.S. consumers use AI to shop, but only 23% trust it to complete a payment.
- Top barriers to AI checkout adoption: spending control, data privacy and dispute access.
- 61% of shoppers trust Visa for agent-initiated transactions, even without a preferred AI tool.
- Retailers can close the gap by setting spending limits, displaying trust signals and phasing in autonomous checkout.
Shoppers Trust AI to Browse, but Not to Buy
New Visa-backed research shows AI is already embedded in the shopping process, but consumer trust drops sharply at checkout. The Harris Poll surveyed 2,065 U.S. adults in late May 2026 and found that 72% have used an AI assistant for product discovery, price comparisons or retailer selection. That number drops to 23% when it comes to trusting AI to complete a payment.
The gap reflects a practical concern. Browsing with AI is low-stakes and easy to reverse. A completed transaction is not. Once a card is charged, a mistake can mean a missed cancellation, an incorrect item or a drawn-out dispute. That’s when brand recognition and consumer protections carry real weight: 61% of respondents said they would trust Visa with agent-initiated transactions, even if they didn’t care which AI tool helped them shop.
The same pattern holds globally. Visa’s Asia-Pacific study, which surveyed 14,764 consumers across 14 markets, found 74% already use AI to shop, while trust and transparency concerns slow the moment of payment. Affluent and digitally savvy consumers show the highest expectations for data control, a useful benchmark for product design and messaging.
What’s Stopping Shoppers from Paying Through AI?
Consumers aren’t rejecting AI shopping outright. They want to know who stands behind the transaction, what the assistant can do and how to fix a mistake. Three concerns drive hesitation: loss of control over spending, uncertainty about data use, and doubt about dispute and refund access.
Competing payment rails, including stablecoins and other faster-settling methods, won’t resolve these concerns on their own. The research suggests most shoppers won’t trade familiar protections for marginal speed or cost gains. Trust is the barrier, not the technology.
How Can Retailers Close the Gap Between Browsing and Paying?
Retailers don’t need to own the AI assistant to win the order. They need to make the handoff from discovery to checkout feel safe, controlled and reversible. Several steps move the needle:
Set clear permissions and spending limits. Let shoppers define a maximum spend that resets on a schedule they choose, preapprove specific retailers or categories and require step-up authentication for anything unusual. Confirm every setting by email or text. Show a change log. Keep the language plain.
Build authentication that stops fraud without adding friction. Support passkeys on web and app. Offer one-time codes via push or SMS. Let low-risk, repeat purchases flow through without extra steps while flagging unusual behavior in real time. Tokenize stored credentials to limit exposure if something goes wrong.
Make receipts and reversals simple. Send a real-time receipt that names the AI channel, the item, the price, the ship-to address and the return window. Include a single-click link to cancel or start a return before the item ships. Publish refund timelines by payment method.
Display trust signals where they count. Show recognized payment brand marks at checkout and in account settings, not just the footer. If you support tokenization or passkey authentication, say so in plain language. Remind shoppers of zero-liability policies and explain how disputes work before they pay.
Explain data use clearly. Spell out what data trains recommendations, what triggers fulfillment, and what never leaves your systems. Offer a privacy summary with toggles for ad use and assistant permissions. Let shoppers opt out of automated payments while keeping AI-assisted browsing on.
Phase autonomous checkout gradually. Start with replenishment orders and low-ticket items. Let the assistant build carts or monitor prices before it can complete purchases. Apply per-order and monthly spending caps, then expand permissions only after a clean track record. Treat it like a loyalty tier: trust earned, controls widened.
Train your service teams to answer one question quickly: what happens if the assistant buys the wrong thing? A clear path to cancellation, receipts and refunds turns a possible friction point into a reason to come back.
(Note: AI assisted in summarizing the key points for this story.)