Agentic commerce

OpenAI adds card payments to ChatGPT dots agents

ChatGPT Wallet lets dots agents complete card purchases after user approval. The first rollout is for US ChatGPT Pro subscribers aged over 18.

Article Image Alt

CNBC reported that OpenAI announced ChatGPT Wallet on 13 November 2026, a payment feature that lets its dots agents make purchases for users after approval. The first rollout is limited to ChatGPT Pro subscribers in the United States who are over 18. The feature brings card networks into the agent workflow, with tokenised payment details rather than raw card data exposed to the agent.

How the payment works

OpenAI built ChatGPT Wallet in partnership with American Express, Mastercard and Visa. Eligible users enter their card information once, then ask a dot to help find items and complete a purchase. Each purchase requires the user to review the order and approve it with a security check such as Face ID or a fingerprint.

OpenAI says its dots agents do not receive the user's actual card details. The payment network provides a payment token for the transaction, including a scoped card number, security code and expiration date. The merchant receives the information needed to fulfil the order, such as shipping address and contact information.

Cardholders keep applicable credit card rewards and protections when they pay through Wallet. OpenAI's terms leave deliveries and returns with merchants, while users generally go to their card issuer for billing disputes. The feature uses existing card payments, not cryptocurrency.

Launch limits and planned controls

Wallet is available in dots conversations, not standard ChatGPT chats. Every purchase currently requires user approval. OpenAI says it is not taking a cut of transactions at launch.

The company plans to add more user controls later, including spending limits and parameters for purchase categories or time periods. A date for those controls has not been announced. OpenAI also wants Wallet to work across more supported retailers.

The limited launch also gives OpenAI and its payment partners a contained environment for testing. ChatGPT has more than 1.2 billion weekly active users and nearly 100 million paying users as of the end of the third quarter, according to OpenAI. A wider release would therefore have operational consequences for merchants, payment processors and fraud teams.

Merchant readiness

ChatGPT Wallet follows OpenAI's earlier commerce work. The company introduced Instant Checkout with Stripe in September 2025, which relied on participating merchants. By March, it had rolled out a shopping experience focused on finding and comparing products.

Wallet shifts the emphasis from product discovery to delegated checkout. OpenAI also published tools and guidance for businesses alongside the announcement, including advice on ensuring that websites are not unintentionally blocking OpenAI agents.

That point matters because some large merchants and service providers do not want agents transacting on their sites. United Airlines has said it will block agents from accessing its systems to book tickets. Amazon prevents agents from purchasing products on its platform.

For enterprises, the same issues appear in a different form. A purchasing agent needs scoped authority, clear user consent, transaction records and fraud controls that recognise delegated action. The technical design around tokenisation does not remove the policy question of when software may spend on behalf of a person.

When an agent can select goods, pass a tokenised payment credential and hand an order to a merchant, the control model moves from chat output to agentic AI, which ECS's Core AI topic covers alongside Azure AI Foundry, the EU AI Act and responsible AI. Sessions and workshops give attendees room to compare practitioner experience on approvals, audit trails, delegation and risk controls. Bring the scenarios where an agent is expected to act, spend or represent a user, because those details decide whether the design is acceptable.


This article was written by the ECS editorial team with the assistance of Amal Agent, run.events' AI assistant, which supported the research, drafting and proofreading.