AI shopping agents between acceptance and mistrust

Many users can easily picture an AI assisting with online shopping by comparing prices and suggesting products – and plenty already use AI for exactly that. But when it comes to trusting an AI assistant with the entire purchase, including payment, hesitation grows considerably.

That is – put somewhat pointedly – the current mood among consumers regarding AI-based shopping assistants. This tension, between the desire for convenience and the fear of losing control, currently defines agentic commerce, the new evolutionary stage of e-commerce.

Interest is real and growing fast. According to the 2026 Mastercard Payment Pulse study, conducted with GfK, a third of German online shoppers already use AI-based features when shopping – among Gen Z, that figure rises to 58 percent. Three-quarters of respondents expect AI agents to become standard in online retail, and among 30- to 39-year-olds, that share reaches as high as 87 percent.

A similar picture emerges from the ECC Köln study "Social, Assisted, Agentic": 61 percent of the roughly 1,000 respondents could see themselves using AI shopping agents – above all for price optimization, travel planning, or buying clothes.

So broad acceptance of this new technology exists – in theory. But what happens when it comes to practical implementation: when AI assistants actually decide independently, on behalf of consumers, which product to select, resulting in a payment-triggering purchase?

Once money and autonomy enter the picture, the mood among respondents shifts.

A YouGov survey conducted on behalf of ACI Worldwide among more than 2,000 UK adults shows:

• Only 19 percent trust an AI assistant with everyday purchasing decisions.
• That compares with 55 percent who trust a human advisor.
• Six in ten respondents would abandon an AI shopping agent after just one mistake.
• And only 18 percent believe an AI would act in their financial best interest.

Particularly telling: 59 percent say they would not trust any organization – neither a bank, a retailer, nor a technology provider – to responsibly manage AI-powered shopping and payments.

The study by the Institute of Retail Management (IRM-HSG) at the University of St. Gallen, conducted in Switzerland, Germany, and Austria, reveals the same fault line. Roughly 15 percent of the nearly 3,000 respondents have already used an AI agent to support an online purchase, but only 5 percent let it handle the entire purchasing process independently. Respondents are happy to delegate price comparisons – almost half would let an AI search for the best deal – but want to keep the final decision, the product recommendation, and the payment step for themselves. A study by ECC Köln arrives at a similar conclusion: only 9 percent would leave a purchase, including payment, entirely to an AI agent.

This reluctance is not an isolated retail quirk – it's part of a broader erosion of trust in AI documented by Fractl's 2026 study among US consumers. A year ago, 82 percent of respondents rated AI as more helpful than traditional search. By mid-2026, that figure has dropped to just 54 percent, and the share who consider AI less helpful has grown sixfold.

The Fractl data also show that heavy, undisclosed AI use is starting to cost brands credibility. The share of respondents who say their trust in a favorite brand would decline as a result has nearly doubled in a single year. And notably, the most critical group is Gen Z – the most AI-savvy age group of all. For retailers and manufacturers weighing how visibly to rely on AI, this is a signal that they should approach AI use deliberately and transparently, rather than focusing solely on efficiency gains.

The same pattern shows up across the studies: shoppers are happy to delegate the tedious, low-risk parts of shopping – price checks, deal-hunting, product research – but resist handing over judgment and money to a system they can't truly hold accountable.

Still, the data suggests the potential is there: even among non-users, 63 percent expect to use AI systems in the future, according to ECC Köln. Current reluctance, then, is better read as "not yet" rather than "never." For now, though, technological development is outpacing consumer trust – and that gap is, to some extent, self-inflicted: according to Fractl, only 20 percent of companies consistently label AI content, even though around 85 percent of consumers expect exactly that.

Building trust in AI use throughout the shopping process is therefore a task for everyone involved – above all technology providers, but also retailers and brands. Concretely, that means: at least in this early era, companies should work with explicit approvals rather than blanket authorizations, along with firm spending limits. Protected payment data and clear accountability when something goes wrong are equally essential. And it means establishing transparency and control as standing rules within organizations – from AI labeling and human review to guaranteeing that customers can reach an actual person when something goes wrong.