ChatGPT Can Now Shop and Pay as OpenAI Enters Agentic Commerce

Share On:

ChatGPT Can Now Shop and Pay as OpenAI Enters Agentic Commerce
In This Article

The new payment feature allows AI agents to complete online purchases after user approval, bringing OpenAI closer to turning conversational AI into a commercial assistant.

Agentic commerce is online shopping in which an AI agent does more than recommend products. It also helps complete the purchase on the shopper’s behalf. On November 13, 2026, OpenAI moved ChatGPT into this territory by introducing ChatGPT Wallet, a payment feature that lets its AI agents, called dots, select products and pay for them once the user approves.

If you searched for whether ChatGPT can now shop and pay, the short answer is yes, with limits. The feature is available only to eligible ChatGPT Pro subscribers aged 18 and above in the United States. Every purchase still needs the user’s approval.

ChatGPT Can Now Shop and Pay as OpenAI Enters Agentic Commerce

OpenAI has taken ChatGPT into checkout, the final stage of the online shopping journey. Until now, AI assistants have mostly helped shoppers research products, compare prices and narrow down choices. With ChatGPT Wallet, OpenAI wants its agents to take part in the transaction itself.

The first release is deliberately narrow. Eligible ChatGPT Pro subscribers aged 18 and above in the United States can use it through dots, and each user needs an eligible US-issued payment card. American Express, Mastercard and Visa are part of the initial rollout, and OpenAI’s updated support documentation also lists eligible Chase cards.

How ChatGPT Wallet handles payments

ChatGPT Wallet is built so that the user stays involved in every purchase. An AI agent cannot spend money without the user knowing.

The process works in four steps:

  1. The user adds an eligible payment card through a secure setup process.
  2. The user asks a dot to find a product or prepare an order.
  3. Before checkout, the agent shows the purchase details, including the items, the total amount and the selected payment card.
  4. The user reviews the order and approves it with a security check, such as Face ID or fingerprint verification.

After approval, the payment is processed with a token generated by the payment network. The token stands in for the user’s actual card details, so neither the agent nor the merchant receives the original card information during the payment.

Purchases can also keep eligible card rewards and existing protections for qualifying disputes. The merchant remains responsible for pricing, delivery, returns and customer service.

Editorial analysis: The approval step is the most important design choice in this launch. A biometric check at the moment of payment gives the user a clear point of consent, and it gives the card issuer evidence that the cardholder authorised the order. That evidence matters if a dispute follows. Based on the details OpenAI has published, the agent prepares the order, but the human authorises the payment. Shoppers should still read the summary screen carefully. The agent chooses the product, so the user is the last person able to catch a wrong size, a wrong seller or an unexpected shipping fee.

Why OpenAI is entering agentic commerce

The move reflects a wider change in how people shop online. Consumers increasingly use AI tools to discover products and weigh their buying options. OpenAI says 1.2 billion people use ChatGPT each week, and many of them already turn to it for product research and purchasing decisions.

Helping someone choose a product and actually buying it are two different responsibilities. Buying requires access to payment infrastructure, reliable product information and a system that can handle transactions securely.

This is where the wallet becomes commercially significant. OpenAI is connecting its AI agents with established payment networks instead of trying to replace the systems retailers already use. That approach lowers the barrier for merchants, because they can accept an agent-led order through the same rails as any other card payment.

OpenAI is also working with payment service providers and fraud-prevention businesses, including Adyen, Stripe, Forter, Riskified, Sardine and Signifyd. The aim is to help legitimate agent-led transactions pass existing payment and risk checks.

Editorial analysis: Fraud systems are usually trained to treat automated behaviour as a warning sign. An agent that logs in, fills a basket and pays in seconds can look like a bot attack. Bringing in the fraud-prevention companies early suggests OpenAI understands that checkout success will depend as much on risk teams as on the AI itself.

Retailers face a new challenge

For online retailers, AI-driven purchasing raises an operational question: are their websites ready to work with automated shopping agents?

OpenAI says some websites and payment systems may unintentionally block agent-led purchases. It has published a Commerce Partner Playbook that offers guidance on three things: making product information accessible, allowing agents to interact with websites, and helping payment systems recognise legitimate transactions.

Retailers also need to decide how much control they are willing to hand to automated systems over the shopping experience. If customers increasingly discover and buy products through AI assistants, businesses may have to rethink how they keep a direct relationship with shoppers.

Editorial analysis: A retailer reviewing its readiness could start with four practical checks:

  • Product data: prices, stock levels, sizes and delivery times should be accurate and readable by machines, since an agent can only act on what it can read.
  • Bot rules: security settings that block all automated traffic may also block legitimate agent purchases.
  • Payment and risk settings: fraud rules should be tested against tokenised, agent-led orders.
  • Customer ownership: the retailer should confirm where the customer’s email, order history and loyalty data end up when the sale happens inside a chat.

The last point is the one most likely to be overlooked. A sale completed inside an assistant may leave the merchant with the revenue but a thinner customer relationship.

What comes next for AI-powered shopping?

The current version still requires users to approve individual purchases. OpenAI has outlined a longer-term plan in which users could set spending limits, choose permitted purchase categories and add time-based conditions. This would let agents handle certain transactions more independently under predefined rules.

For now, availability is limited to eligible US users, and the feature is not a universal checkout option across all retailers.

The bigger story is the direction of travel. AI assistants are moving from answering questions and recommending products toward carrying out tasks that have direct financial consequences. Whether consumers accept this model will depend on three things: how reliably the systems work, how clearly they communicate purchase details, and how well payment safeguards protect users.

Editorial analysis: The promised spending limits and category rules will be the real test. Approving each purchase is simple to understand. Handing an agent a monthly budget for groceries is a much bigger act of trust. The first mistakes in that setting, such as a duplicate order or a wrong substitution, will shape public confidence more than any launch announcement.

Author:

Related Posts