If Checkout Pages Go Away, Fraud Systems Have to Change What They Ask

Stripe's president, William Gaybrick, has said that even a less ambitious version of agentic commerce could mean the end of checkout pages as we know them. If an AI agent completes the purchase, behaviour that once looked automated - and therefore suspicious - may be a legitimate order. Fraud systems will still need to know whether the payment method is legitimate, valid, and authorised by the cardholder. The harder question becomes whether the machine acting for the customer can be trusted, and which signals survive if the checkout page disappears.

Stripe's president says checkout pages will go away as AI commerce expands.

William Gaybrick said that even a less ambitious version of agentic commerce could mean the end of checkout pages as we know them. This might be one of the biggest changes coming to ecommerce.

Instead of a customer finding a product, moving through checkout, and manually entering their information, an AI agent could increasingly complete that process on their behalf.

Legitimate purchases may look automated

This creates a very interesting problem. For years, fraud systems have been built around determining whether a transaction looks like legitimate human behaviour. Typing patterns, mouse movements, session flows, IP addresses, and other behavioural signals can all contribute to understanding whether the person behind a transaction looks legitimate.

What happens when the legitimate customer isn't the one navigating the purchase? The behaviour that might historically have looked automated - and therefore suspicious - could represent a completely legitimate purchase. That means one of the questions fraud systems ask may need to be: is this legitimate machine behaviour?

Checkout data may disappear with the checkout page

There's another challenge here too. If the checkout page disappears, the data we currently collect during checkout could disappear with it. Signals that have historically been useful for risk evaluation may no longer exist in the same form.

Fraud systems will need to understand:

• Which existing signals remain useful
• Which behavioural signals lose relevance
• What new signals can establish whether an agent is legitimate
• How to distinguish authorised agents from malicious automation
• How to compensate for data that disappears with the traditional checkout journey

One fundamental question doesn't change. Whether a human completes the checkout or an AI agent does it for them, we still need to understand whether the payment method is legitimate, valid, and authorised by the cardholder.

The future of fraud detection may be less about asking whether there's a human behind the transaction, and more about whether the machine acting for them can be trusted.

Related: fraud trends 2026, how it works, pricing, and the FAQ.

If you want ranked rule changes from the transaction and chargeback data you already have - book a demo.

Originally shared on LinkedIn.