Nearly three out of four consumers worldwide now use artificial intelligence in some part of their online shopping experience, according to a new survey released Monday, marking a major shift in e-commerce behavior with potentially serious implications for fraud and liability.
The findings, based on a global survey of more than 5,000 consumers, point to the rapid rise of so-called “agentic shopping,” where AI tools assist in browsing, comparing and in some cases even completing purchases.
While only 13% of respondents said they had completed a purchase referred by an AI assistant, 70% reported being at least somewhat comfortable allowing an AI agent to make purchases on their behalf. More than half said they plan to use such tools for holiday gift shopping this year.
The survey, released by e-commerce risk management firm Riskified, highlights both the convenience and the vulnerabilities associated with AI-driven shopping. Shoppers are already using tools like ChatGPT to generate product ideas (45%), summarize reviews (37%) and compare prices (32%). But the growing reliance on AI also blurs lines of accountability when things go wrong.
“AI shopping agents may make buying easier for consumers, but they also blur the lines of accountability for fraud and policy abuse,” said Jeff Otto, Riskified’s chief marketing officer. He warned that merchants may still be held liable for disputed transactions, even when a customer never directly interacted with their site.
According to the company, bad actors are likely to exploit the AI shift by targeting weaknesses in identity verification, payment systems and customer service channels. As AI automates more of the shopping process, these gaps could leave businesses vulnerable to a rise in chargebacks and other forms of fraud.
Consumers surveyed also expressed concerns about the AI shopping experience. The top worries included payment security (32%), privacy (26%), potential mistakes (18%) and loss of control (17%). Still, trust in AI is growing: 36% of respondents said they trust AI to influence their purchases, compared to 38% who trust in-store associates. Only 25% preferred to shop online without AI assistance.
The release of the survey comes amid the rollout of the Agentic Commerce Protocol (ACP), an emerging open standard designed to enable purchases directly through AI interfaces. Riskified recommends that e-commerce companies adapt by investing in fraud detection tools, demanding more data transparency from AI platforms and using shared intelligence to regain lost context in agentic transactions.
Otto added that fraud prevention teams are uniquely equipped to help businesses adopt AI shopping safely: “Their expertise isn't about saying 'no' to innovation; it's about building the guardrails that allow the company to say 'yes' confidently. They are essential to ensuring that exciting new channels like this become sustainable revenue drivers, not vectors for fraud and abuse.”
The survey was conducted online between September and October 2025, polling 5,400 adults ages 18 to 64 in markets including the United States, United Kingdom, Brazil, Mexico, Singapore, Australia, China, the Netherlands and the United Arab Emirates. All respondents had made at least one online purchase in the prior three months.