Copy article

58% of UK online retailers say AI agents are already buying on their platforms.

ended 22. May 2026

The Retail Gazette reports The Payments Association surveyed 100 senior finance, payments and risk leaders at UK online retailers during Q1 2026. More than half said AI-initiated transactions have already reached their systems. Not one called agentic commerce irrelevant. Yet when presented with a disputed £2,000 AI agent purchase, no majority position on liability emerged, 24% said it depends, 21% wanted shared liability, 18% pointed at the AI vendor. Only 41% felt confident in current frameworks.

Savvy shoppers don't hand their needs to an AI agent because they love automation. They do it because online shopping has become real work: opaque pricing, dark patterns, cookie banners, SEO-stuffed product descriptions and other friction retailers built in to suit their needs. An agent cuts through all of that.

UK Strong Customer Authentication rules under PSD2 assume a human responding to a real-time challenge. An AI agent on delegated instructions breaks that consent model. The Payment Services Regulations 2017 require immediate refunds for unauthorised transactions but don't allocate liability between payment providers and AI technology providers.

Fraud engines trained on human behaviour are flagging legitimate agent traffic as suspicious. Visa, Mastercard, Google and OpenAI have published agentic commerce protocols. 

We'd like your views:

  • If AI agents always surface the cheapest price, does brand loyalty survive? Or do retailers end up competing on margin alone while consumers pocket the savings, regardless of who makes the click?
  • When an AI agent makes a disputed purchase, current frameworks assume a human made the decision. Should consumers lose dispute rights for agent errors, or does that punish people for using the type of tools the industry uses against them?
  • Fraud engines trained on human behaviour are flagging legitimate agent traffic as suspicious. At what point do false declines cost retailers more than the fraud they were built to catch?
  • The FCA says it's considering whether regulation needs to change. Merchants are already processing agent transactions they can't identify. Is big business the one to lose for operating slower than the customer in the market?

3 responses from the Newspage community

Copy all

Star Quote
Copy

Colette MasonVerified
AI Ethics Consultant at Clever Clogs AI
Forget the philosophical debate about AI personhood in the eyes of the law. Right now, forward-looking customers don't care. They've worked out that an agent scanning every competitor's price in seconds gives them leverage no loyalty scheme can match. That's not fraud, it's rational shopping.

I find it ironic that the same AI-research power letting a legitimate agent find the best deal for a shopper is now frowned upon by high-tech retailers. Are retailers ready to compete honestly on price when the customer finally has as much market information as the seller?

Retailers built self-interested friction into their platforms to protect margins: dark patterns, obstructive comparison, buried pricing. Consumers are battle-fatigued.

However, that same AI-power also lets bad actors exploit weak authentication, initiate disputed transactions and game refund processes at scale. Tumbling retail profits means the fraudulent transactions slipping through the cracks are a problem for all of us.
Copy

Editor
Kate AllenVerified
Owner at Finest Stays
Whilst we're not yet seeing AI agents directly making bookings through our website, it does make me very uncomfortable. For higher-value purchases like holidays, we need absolute certainty that a booking has been intentionally made. In my view, the risk should sit with consumers and the tools they choose to use, not businesses trying to work out whether an AI genuinely acted on someone's wishes.
Copy

Adam Baldwin0
Tech & AI Legal Strategist | DE&I | AI Governance & Digital Transformation at Moloco / She Shapes Tech / City of London Corporation
There is also a fundamental fairness issue. Retailers, payment providers and platforms already rely heavily on AI for pricing, targeting, fraud scoring and behavioural optimisation. It is difficult to justify a framework where businesses can use automation to influence consumer decisions, while consumers lose protections for using automation to navigate the same market more efficiently. From a regulatory perspective, particularly under the FCA’s Consumer Duty regime, penalising consumers for adopting mainstream AI tools would likely be viewed as an unfair allocation of risk rather than a legitimate fraud-control measure.