When AI shops on your behalf, who owns the transaction?

The next evolution of digital commerce may not be driven by consumers clicking ‘buy now’, but by AI acting on their behalf. According to dispute specialist Chargebacks911, that shift could herald a new and more complex wave of payment disputes.

AI – who owns the transaction?

Monica Eaton, the company’s founder and chief executive, argues that the industry is approaching an inflection point.

As AI systems move beyond recommending products to autonomously completing transactions, traditional definitions of customer intent are being tested.

“We are about to see a different type of chargeback,” she warns. The payment may be authenticated, the merchant error-free, and the cardholder account uncompromised. Yet the customer may still claim: I did not want that purchase.

This emerging friction sits at the heart of what has become known as agentic commerce.

Rather than serving as a passive assistant, AI can renew subscriptions, re-order household goods, switch brands to save money, or book travel based on diary availability.

From a system perspective, the parameters may have been followed precisely. From a human perspective, however, the outcome may feel misaligned with personal preference or shifting circumstances.

The Disappearance of the Click

For decades, payments infrastructure has treated the consumer’s click as the ultimate proof of intent. Authentication checks, two-factor verification and digital signatures are all designed to demonstrate that a human actively authorised a transaction.

Agentic commerce disrupts that model. When an AI agent executes a purchase within pre-approved boundaries, there may be no immediate, conscious act by the consumer at the point of sale.

The evidential anchor shifts from a real-time action to a prior consent framework.

Payment networks are already preparing for this transition. Both Visa and Mastercard have begun piloting autonomous, agent-initiated transactions with banking partners.

These initiatives are designed to preserve security and control while enabling machine-driven commerce. Yet they also raise a fundamental question: how should intent be evidenced when the human is not directly involved?

A New Category of Dispute

Traditional chargeback cases rely on familiar proof points: successful authentication, confirmed delivery, or evidence of cardholder participation. In an agent-led model, those markers may be insufficient.

Instead, dispute resolution could hinge on whether the consumer clearly authorised the agent’s scope of action, what spending limits were in place, whether notifications were delivered, and how transparent the transaction trail was.

If background purchases become commonplace, consumers may increasingly encounter charges they do not immediately recognise. The instinctive response, as Eaton notes, is often to dispute first and investigate later.

Compounding this shift is the emergence of outcome-based pricing models, in which platforms levy fees tied to completed AI-driven transactions. Such structures could accelerate the volume of purchases occurring without an obvious moment of human intent.

For merchants, the implications are strategic rather than merely operational. Clear permission architectures, granular controls and robust audit trails will become essential.

Agentic commerce may promise efficiency and higher conversion rates, but unless customer intent remains demonstrably central to the transaction, the industry risks replacing one category of fraud with another: disputes born not of theft, but of algorithmic misunderstanding.