Autonomous agents are now handling purchases. What’s left for PSPs to optimise?
For years, the checkout page has been a focal point of digital commerce. It is where conversions are won or lost, fraud is detected, and value is captured. Now, Google is repositioning it; by removing the interface altogether.
At its I/O developer event earlier this month, Google introduced “agentic checkout” as part of a broader AI-driven shopping experience. The announcement has not come with much fanfare from the payments industry, but its implications are wide-reaching.
At the heart of the feature is a shift in control: transactions can now be initiated, executed and completed by artificial intelligence, with minimal or no user involvement.
The system allows users to set preferences, such as price range, colour, and size, for products they are interested in. Google’s AI monitors these listings, and when criteria are met, it can finalise the purchase on the user’s behalf using Google Pay. There is no basket, no visible form, and no button to click.
Changing the visibility of payments
For payment providers, this introduces a structural change. The transaction no longer occurs at the end of a merchant’s checkout flow, but within Google’s environment. This risks pushing PSPs further into the background. Payments become a function, not a feature.
As the visible components of a transaction disappear, the opportunity for providers to differentiate through user experience, branding or value-added services narrows. The integration layer becomes the battleground — and it is a quieter one.
Limitations of current fraud tools
Most fraud prevention models rely on behavioural patterns that indicate human activity. These include device fingerprints, typing cadence, and purchasing habits. Agentic checkout removes these signals.
An AI agent does not hesitate, misclick, or type irregularly. It follows a script. That uniformity makes it difficult to assess risk using traditional methods. Without clear signals of human intention, providers may struggle to distinguish legitimate agent activity from automation-based fraud.
This will likely require a new class of fraud tools — not simply adapted models, but systems built for an environment where transactions are both autonomous and obscured.
Margins and the role of Google Pay
The decision to route payments through Google Pay raises questions about the economics of agentic commerce. As more purchases take place within Google’s framework, the tech company may be in a position to negotiate reduced transaction fees or steer users towards low-cost rails.
Card networks and PSPs may find themselves excluded from default flows unless they offer something more than processing. Margins, already thin in many parts of the market, could face further pressure.
There is also a shift in the merchant relationship. In agentic checkout, the final interaction does not take place on the merchant’s site. The consumer may never visit the merchant directly, and the purchase may occur without the merchant being part of the customer’s decision-making journey.
This challenges existing partnerships and pricing models. Payment providers who built their value proposition around helping merchants optimise checkout may need to rethink their position in the chain.
Responding to the Shift
Agentic checkout is not a universal standard — yet. It is launching in the United States and will take time to scale.
Payment providers will need to:
- Develop fraud detection systems that account for autonomous agents
- Build APIs and services that integrate smoothly with agentic platforms
- Offer new forms of data reconciliation and transparency for merchants
- Participate in discussions around agent identity and transactional accountability
If the interface changes, so does the infrastructure.
So while Google has not created a new payment method, the alteration of the play where payments occur and who initiates them is significant enough. The checkout experience is becoming less visible, less interruptive and, for some parties, less accessible.
