Visa plugs payment network into ChatGPT, letting AI agents shop and pay
Visa has integrated its payment network directly into OpenAI’s ChatGPT, allowing the chatbot to independently facilitate e-commerce transactions on behalf of users. The collaboration enables AI agents to make purchase decisions and complete payments at any merchant that accepts Visa, moving beyond previous limitations where digital shopping assistants were restricted to specific retail platforms or select partners.
Did You Know? Visa’s previous industry attempts at enabling agent-based shopping were confined to single retailers or small, enrolled groups; this new integration marks a shift toward a broader, network-wide capability.
How the Visa and ChatGPT integration works
The collaboration relies on a division of labor between the two technology firms. OpenAI provides the AI architecture that allows agents to interact with users, evaluate products, and initiate transactions. Visa serves as the infrastructure provider, handling payment authorization, fraud monitoring, and security protocols at scale.
According to Jack Forestell, Visa’s chief product and strategy officer, the system is designed to provide a “superior discovery experience.” A user might ask the chatbot to find wireless headphones under $150, and the agent would locate and purchase the item. To address security concerns, Visa has implemented guardrails including spending limits, mandatory approval steps, and a list of approved merchants.
Why the shift to autonomous purchasing matters
The integration represents a significant departure from OpenAI’s previous e-commerce effort, known as Instant Checkout. That tool, which allowed ChatGPT to act as a personal shopper, was retired in March after failing to gain widespread merchant adoption. Merchants reportedly cited the 4% transaction fee charged by OpenAI as prohibitively expensive.
While Visa and OpenAI have not disclosed the financial terms of this new arrangement, the move carries substantial implications for banking and retail. Financial institutions have expressed concerns regarding fraud and the potential for unauthorized transactions. In response, Visa is updating its token framework and data capture processes under its “Visa Intelligent Commerce” initiative to manage disputes and clarify the chain of responsibility between the consumer, the AI agent, and the merchant.
Expert Insight: The transition from AI-led product recommendations to autonomous execution creates a new challenge for the payment industry. By moving the point of authorization to an AI agent, Visa is attempting to standardize a “trust infrastructure” that protects banks and consumers from the risks of overspending or incorrect orders, effectively treating the AI as an authorized proxy for the account holder.
What may happen next for AI-driven commerce
Market analysts expect that human oversight will remain a primary component of AI-assisted shopping in the near term. Visa anticipates that most initial transactions will continue to require a human to review and approve the purchase after a notification from the agent. Over time, as consumer trust in the underlying security infrastructure grows, it is possible that users will opt to waive these manual verification steps for routine or recurring purchases.
This development sets the stage for increased competition in the payment sector. Mastercard has also begun introducing AI-shopping features, though its current focus centers on business-to-business procurement—such as allowing an AI to authorize payments for advertising services on behalf of a small business.
Frequently Asked Questions
What protections are in place to prevent unauthorized AI purchases?
Visa has stated that the feature includes guardrails such as pre-set spending limits, required user approval steps for transactions, and a curated list of approved merchants.

How does this differ from OpenAI’s previous e-commerce tool?
OpenAI’s previous “Instant Checkout” was limited to select merchants and was widely rejected due to a 4% transaction fee; the new Visa integration is designed to work across any merchant that accepts the Visa network.
Who is responsible if an AI agent makes a mistake?
Visa plans to handle disputes using its existing rules regarding consumer intent and merchant processing. The company is also modifying its token framework to address potential errors that may occur during the automated transaction process.
Would you trust an artificial intelligence agent to manage your routine household shopping without your direct, real-time approval?