OpenAI is moving beyond information retrieval to direct commerce with the launch of ChatGPT Wallet, a system designed to let AI agents execute transactions without ever accessing a user's actual credit card details. By partnering with major payment networks like American Express, Mastercard, and Visa, the company aims to facilitate 'agentic commerce,' where users can delegate shopping tasks to AI while maintaining control over spending and security.

This image illustrates the concept of 'agentic commerce' and the integration of shopping-related tasks into the ChatGPT ecosystem.
The Mechanism of Agentic Payments
The core functionality of ChatGPT Wallet relies on a 'tokenized' payment flow. A user adds their card to the wallet once, and the system handles the heavy lifting of transaction details.
When an agent identifies a product to buy, it requests a purchase, which the user must then review and approve. Upon approval, the payment network generates a payment token that acts as a one-time substitute for the card number, security code, and expiration date.
This allows the transaction to complete through existing checkout flows without the merchant or the AI agent ever possessing the user's primary card information.
“ChatGPT Wallet gives agents a secure way to pay on a user’s behalf. Users add a card once, then approve each purchase without re-entering their payment details.”
— Introducing ChatGPT Wallet
Security and User Protections
To prevent unauthorized spending, every purchase requires a proactive security check from the user. This ensures that while an agent can navigate a website and select items, it cannot finalize a payment without a physical biometric confirmation from the account owner.
OpenAI designed the wallet to preserve existing consumer protections. Users still retain their card's rewards and are eligible for standard dispute protections, such as those for defective items, ensuring that the shift to AI-mediated shopping does not strip away traditional banking safeguards.
Infrastructure and Business Integration
For businesses, the transition to agentic commerce presents technical hurdles. Many websites currently use firewalls or 'Web Bot Auth' systems that might inadvertently block ChatGPT agents.
OpenAI is releasing tools to help businesses recognize signed requests from agents and update their product feeds to ensure accurate pricing and availability.
Because many risk systems are trained to flag automated behavior, OpenAI is working with payment service providers like Adyen, Checkout.com, and Stripe to help their systems recognize legitimate purchases made by agents.
Future Delegation and Limitations
While the current rollout requires manual approval for every purchase, OpenAI intends to move toward full delegation. In the future, users may be able to set specific parameters—such as spending limits, categories, or time periods—to pre-approve a range of purchases automatically.
The success of this system depends on the reliability of the agents in navigating complex web environments. The current rollout is limited to ChatGPT Pro users in the US to allow OpenAI to learn from real-world use and improve compatibility with existing banking and payment infrastructures.
What Hacker News readers are saying
- Experience: Several commenters shared their experience with ChatGPT's current limitations, noting that the model frequently recommends products that are out of stock, unavailable, or have inaccurate pricing. comment ↗
- Skeptical: One commenter argued that most people prefer to conduct their own research and review images before purchasing, suggesting that the subset of people who view shopping as a completely automatable commodity is smaller than tech companies believe. comment ↗
- Skeptical: Another user expressed concern that OpenAI's move to provide its own payment solution could alienate existing payment service providers and position OpenAI as a direct threat to those businesses. comment ↗
- Open question: One commenter raised questions about the downstream effects on the web, wondering if this could lead to more aggressive paywalls or a more 'enshittified' internet as sites seek to monetize AI agent traffic. comment ↗
Summarized opinions from the Hacker News thread, not verified facts. Read the full discussion ↗