ChatGPT Wallet, which OpenAI launched on 13 November 2026, lets its AI agents complete purchases for users, and every purchase requires the user to approve it with Face ID, a fingerprint or a similar security check. The agent pays with a network-issued payment token, so neither the agent nor the merchant receives the user’s real card details.
ChatGPT Wallet Starts With US Pro Subscribers and Three Card Networks
OpenAI built the wallet with American Express, Mastercard and Visa, and it works through dots, the company’s personal agents inside ChatGPT. At launch it is limited to ChatGPT Pro subscribers in the US who are at least 18 and hold an eligible US-issued card, OpenAI says. The company describes the narrow release as a way to learn from real use with businesses, banks and networks before expanding. OpenAI says 1.2 billion people use ChatGPT each week, and that many already use it to research and compare products before buying.
Each Agent Purchase Produces a Biometric Approval Record
A user adds a card once. Before each purchase, the user sees the items, the total and the card that will be charged, then approves with a security check such as Face ID or a fingerprint scan. OpenAI says each approval creates a clear record of the purchase the user authorized the agent to make. The merchant receives what it needs to fulfill the order, such as the shipping address and contact details, TechCrunch’s Sarah Perez reported. Users keep their card rewards and their protections for eligible disputes, and merchants keep control of pricing, fulfillment, returns and customer service.
A Payment Token Keeps the Real Card Number Away From the Agent
Once the user approves, the payment network supplies a token with its own card number, security code and expiration date, and the charge runs against the selected card through that substitute. The agent and the merchant never see the underlying card details. Because the token looks like an ordinary card to the merchant, OpenAI says it works with existing checkout flows.
Merchants Must Configure Firewalls and Fraud Models to Accept Agent Buyers
OpenAI acknowledges that some sites may block purchases through ChatGPT Wallet without meaning to. Its Commerce Partner Playbook tells businesses to configure their CDN or firewall to recognize signed agent requests using Web Bot Auth, and to check that general security rules are not turning ChatGPT agents away. It also warns that risk systems may flag legitimate agent purchases and cause declines or cancellations. OpenAI says it is working with fraud prevention firms Forter, Riskified, Sardine and Signifyd and with payment providers Adyen, Checkout.com and Stripe so their systems recognize these transactions. Some large merchants are moving the other way. TechCrunch notes that Amazon blocks AI agents and steers shoppers to its own Alexa agent, and that DoorDash and Uber have reported low order volumes from agents.
OpenAI Plans Pre-Approved Agent Spending Within Set Limits
OpenAI says it expects to let users delegate purchases entirely by setting spending limits, purchase categories or time periods, with enhanced safeguards built on biometric checks and authorization records. That would move the approval from each transaction to a policy the user signs once. The company has not given a date for that change.
Why This Matters for Engineering Leadership
OpenAI has made the human approval record the core control in agent payments. The agent can browse, compare and fill a cart on its own, and money moves only after a biometric step that leaves a record tied to that purchase. Teams building agents that take consequential actions inside their own companies, such as deploying code, issuing refunds or changing access, can borrow the same structure: let the agent prepare the action, require a strong per-action approval that produces a durable record, and keep credentials in a token the agent never holds. The planned move to pre-approved limits is the harder design problem, because the record then shows a standing policy signed once for many purchases, and limits, categories and expiry become the controls an auditor reviews. For teams that run e-commerce or payments, the playbook is a near-term work item. Bot rules at the CDN and fraud models tuned to human behavior will treat agent buyers as threats unless someone updates them to verify Web Bot Auth signatures and recognize tokenized agent transactions.
FAQ
What is ChatGPT Wallet?
ChatGPT Wallet is an OpenAI payment feature, launched on 13 November 2026, that lets ChatGPT agents called dots complete purchases for a user. It was built with American Express, Mastercard and Visa.
Who can use ChatGPT Wallet?
At launch it is limited to ChatGPT Pro subscribers in the US who are 18 or older and have an eligible US-issued card. OpenAI says it will expand availability after learning from early use.
Does the AI agent see my card number?
No. After approval, the payment network issues a payment token with its own card number, security code and expiration date, so neither the agent nor the merchant receives the real card details.
How is each ChatGPT Wallet purchase approved?
The user reviews the items, the total and the selected card, then approves with a security check such as Face ID or a fingerprint scan. Each approval creates a record of what the user authorized.
Can agents buy things without approval?
Not yet. OpenAI says it expects to let users pre-approve purchases within spending limits, categories or time periods in the future, with added safeguards.
What do merchants need to change to accept ChatGPT Wallet?
OpenAI’s Commerce Partner Playbook tells businesses to let their CDN or firewall recognize signed agent requests through Web Bot Auth and to work with payment and risk providers so legitimate agent purchases are not declined.
