ChatGPT Wallet: When an Agent Checks Out, Is Your Store Ready?
As of this week, a ChatGPT agent in the US can finish a checkout and pay, and the store never sees the buyer's real card number. OpenAI switched on ChatGPT Wallet on Friday with Visa, Mastercard and American Express live at launch, for Pro users to start. The user adds a card once and approves each purchase with a fingerprint or Face ID. The network then issues a one-time token that settles against the real card, with rewards and dispute protections intact.
The work lands on merchants. OpenAI is asking stores to configure their firewalls to recognize signed agent requests, because standard bot defenses will read a paying agent as an attacker and block the sale. Fraud vendors are in the loop from launch (Forter, Riskified, Sardine, Signifyd), but a store still has to decide to be shoppable by machines.
Ad creative still persuades a human, and that still holds. What changed is who completes the purchase. The human approves once, and an agent carries the order the rest of the way, so a campaign can now win attention and then lose the sale to a bot filter. That leak sits after the ad, in infrastructure most D2C teams have never audited.
D2C founders: if an agent tried to buy from your store today, would the order go through, or get flagged as a bot?
Try our free ad tools: https://igniteai.in/tools/
Sources
First published on our LinkedIn page on 15 November 2026.
Try it on your own product
IgniteAI makes AI video ads for small brands. Script and pictures free; pay only when you are happy with them. No card to start.
Make my first ad →