OpenAI's ChatGPT Wallet Lets Dots Shop And Check Out

OpenAI launched ChatGPT Wallet on Friday, a feature that lets ChatGPT subscribers hand purchasing decisions to AI agents, called dots, and lets those agents complete the checkout. Availability is limited to a small group of U.S. ChatGPT Pro subscribers, age 18 and up, who hold a U.S.-issued Visa, Mastercard or American Express card.

That narrow launch matters more than the announcement itself. This is the first checkout agent from a frontier lab operating on top of existing card rails, rather than a retailer selling its own assistant.

How does a dot actually check out?

Setup happens once. Subscribers add an eligible card to ChatGPT Wallet, then ask their dot to help with a purchase. Payment data is stored, so users do not re-enter card details for every transaction.

At checkout, the wallet charges the selected card using a payment token, a substitute for the actual card number. OpenAI says that token keeps the card information from reaching either the dot or the merchant. The merchant still receives what it needs to fill the order, including shipping address and contact information, sent by the dot.

OpenAI says users are "fully in control of the purchases," reviewing the items, the total and the card before the charge, and approving with a security check such as Face ID or a fingerprint. That approval creates a record of the transaction.

The money path does not change. Customers keep earning card rewards and keep the protections that come with the card, including the right to dispute a transaction. No new payment network appears here. An agent simply occupies the position a shopper's browser used to hold.

Who gets access, and what comes later?

Limited availability at launch lets OpenAI test and improve the feature before a wider release, the company says. More control features sit further out: spending limits, plus parameters for purchase categories or time periods, so users can pre-approve what their agent handles autonomously. That is a roadmap, not a shipping capability.

The competitive evidence is mixed, and OpenAI is not the only claimant. Meta's Muse, which now has over 10 million users, has seen AI-assisted shopping become a top activity on its app. Other personal agents, such as Instinct, already support purchases on users' behalf. Against that, DoorDash and Uber have so far reported low numbers of orders placed by AI agents. One consumer AI app with a large installed base is converting shopping intent; two delivery platforms are not.

Why does Amazon decide the outcome?

OpenAI aims to make the feature work at a broader number of supported retailers. That is the hard part. Some retailers block AI agents from using their sites, and the largest retailer to do so is Amazon, which wants to steer shoppers to its own agent, Alexa. A wallet cannot route traffic to a merchant that refuses the agent.

OpenAI published a Commerce Partner Playbook with tools and guidance for sites that want to serve agent-driven customers. It is an attempt to convert blockers into partners.

What should a chip supply chain reader take from this?

Little, for now, and the absence is the point. The announcement names no accelerator vendor, reserves no inference capacity, and carries no wafer commitment or process node. It offers no transaction count, no merchant list, no take rate, and no revenue estimate from any analyst.

The implied workload is small: a short inference for product selection, then an API call to a checkout service. The honest read is that agentic commerce will pull inference demand, but this release proves nothing about silicon.

What happens next?

The near-term milestone OpenAI describes is spending limits plus category and time-based pre-approval. If that ships alongside a broad merchant list, ChatGPT Wallet stops being a pilot and starts competing with Alexa for the checkout layer, which is where the margin lives.