Do you want me to just not ask?
In a move that tests the boundaries between human intention and machine action, Visa has embedded its payment infrastructure into ChatGPT, allowing an artificial intelligence to shop, decide, and buy on a person's behalf across the full breadth of the Visa merchant network. The partnership, announced in San Francisco, marks a departure from earlier, narrower attempts at AI-assisted commerce — ones that stumbled on merchant resistance, technical error, and the quiet but persistent question of trust. What Visa and OpenAI are building is not merely a faster checkout, but a new kind of relationship between a person's wallet and a machine's judgment.
- The stakes are high: Visa is wagering that consumers will surrender one of their most deliberate daily acts — spending money — to an AI agent they cannot fully observe.
- Previous attempts at autonomous AI shopping collapsed under their own weight, with OpenAI's own Instant Checkout shutting down in March after merchants balked at a 4 percent transaction fee and the system proved unreliable.
- Visa is constructing guardrails — spending caps, merchant whitelists, mandatory approval steps, and redesigned fraud detection — to give consumers enough confidence to take the first step.
- The financial terms between Visa and OpenAI remain undisclosed, leaving the question of merchant tolerance unresolved and echoing the very friction that doomed earlier efforts.
- The trajectory points toward full autonomy: once a user approves the same AI purchase a thousand times, the system may quietly ask whether it should simply stop asking.
Visa has integrated its payment network directly into ChatGPT, enabling the chatbot to browse the internet, identify products matching a user's criteria, and complete purchases autonomously across any merchant that accepts Visa cards. The announcement, made at a company event in San Francisco, signals a meaningful shift in what AI shopping assistants are capable of — and what they are being asked to become.
Earlier attempts at automating purchases were narrow by design. Amazon's Alexa could only shop within Amazon. OpenAI's own Instant Checkout feature, launched late last year, let ChatGPT hunt for products online, but it was error-prone and charged merchants a 4 percent fee they were unwilling to absorb. It was shut down in March. Visa's model is structurally different: users link their Visa cards to ChatGPT, OpenAI provides the intelligence to interpret preferences and initiate purchases, and Visa supplies the authorization and fraud detection infrastructure to operate at scale.
Jack Forestell, Visa's chief product and strategy officer, illustrated the vision simply — a user asks for wireless headphones under $150, and the AI finds and buys them. But he acknowledged that the leap from recommendation to autonomous transaction demands a kind of trust most consumers do not yet hold. To cultivate it, Visa is installing spending limits, approval requirements, whitelisted merchants, and a redesigned token and data-capture framework to handle disputes when something goes wrong between a willing buyer and a compliant seller.
The financial terms of the partnership were not disclosed, a silence that carries weight given how merchant fees ended the last experiment. For now, most transactions will still require a human confirmation step — the AI notifies, the user approves, the purchase completes. But Forestell gestured toward the longer horizon: a future in which repeated approvals make the question itself obsolete. Mastercard is pursuing a narrower version of the same idea, focused on business procurement rather than consumer retail. The infrastructure for autonomous AI commerce is being assembled. Whether people will choose to live inside it remains the open question.
Visa has woven its payment network directly into ChatGPT, betting that people will soon feel comfortable letting an artificial intelligence agent handle their shopping. The payments giant announced the integration Wednesday, saying it now allows the chatbot to browse the internet, find products that match a user's specifications, and complete the purchase without human intervention—all across any merchant that accepts Visa cards.
This represents a significant expansion of what AI shopping assistants can do. Previously, when payment networks or tech companies tried to automate purchases, they were confined to single retailers or small groups of enrolled merchants. Amazon's Alexa could only shop on Amazon. OpenAI itself launched a feature called Instant Checkout late last year that let ChatGPT hunt for specific items online like a digital personal shopper, but the system was error-prone and merchants resisted it because OpenAI charged them 4 percent of each transaction's value. The company shut down Instant Checkout in March.
Visa's approach is different. Users will link their Visa cards directly to ChatGPT. OpenAI provides the artificial intelligence that lets agents understand what a person wants, make decisions, and initiate purchases. Visa supplies the payment authorization and fraud detection infrastructure to make this work at scale. Jack Forestell, Visa's chief product and strategy officer, described a simple scenario at a company event in San Francisco: a customer tells ChatGPT they want wireless headphones under $150, the chatbot finds a pair and buys it. The leap from recommendation to autonomous purchase, Forestell acknowledged, requires a different kind of trust than most people currently have.
That trust has to be built on infrastructure. Visa is installing guardrails: spending limits, required approval steps, and whitelisted merchants. The company is also modifying how it captures transaction data and handles disputes. When something goes wrong—the wrong item arrives, a customer overspends, or fraud occurs—Visa says it will apply the same logic it uses for any other transaction: Did the consumer genuinely intend the purchase? Did the merchant process it correctly? But Forestell noted that the framework may need to shift when both the consumer and merchant did everything right but something failed in between. That's why Visa is redesigning its token framework and data capture process.
Neither Visa nor OpenAI disclosed the financial terms of their partnership or what fees customers and merchants would pay. That silence matters. The 4 percent fee that killed Instant Checkout showed that merchants have limits on what they'll tolerate. Forestell acknowledged that widespread adoption will take time. Most transactions initially will still require human approval—the AI sends a notification, the customer confirms, and the purchase goes through. But he hinted at the long game: once a person approves the same type of transaction a thousand times, the agent might ask, "Do you want me to just stop asking?"
Mastercard, Visa's largest competitor, is also moving into AI shopping, though on a smaller scale. It announced that AI agents can procure services on behalf of businesses—a coffee shop, for example, could authorize an agent to buy advertising services from multiple providers to launch a campaign. The difference is scope: Visa is aiming for consumer shopping across the entire Visa network, not just business procurement or single-retailer scenarios. The question now is whether consumers will actually want this. The infrastructure is being built. The question is whether people will trust it.
Notable Quotes
As AI agents become active participants in the economy, Visa's focus is to ensure transactions are trusted, secure and seamless.— Jack Forestell, chief product and strategy officer at Visa
Making the leap from having AI agents recommend what to buy to doing the purchasing just requires a whole different level of trust.— Jack Forestell, Visa
The Hearth Conversation Another angle on the story
Why does Visa think people will accept this? Autonomous shopping feels like a big leap from recommendation.
Forestell was honest about that. He said people are already comfortable with AI finding things for them—that's become a superior discovery experience. But buying on their behalf is different. It requires trust in the underlying infrastructure, the security, the rules. That's what Visa is trying to build.
What went wrong with OpenAI's Instant Checkout?
Two things. The system made mistakes—it wasn't reliable. And merchants hated the 4 percent fee. It was too expensive for them to accept. Visa hasn't said what it will charge, which is telling. That fee structure could make or break this.
So how does Visa prevent fraud or mistakes?
Spending limits, approved merchant lists, and approval steps. At first, most transactions will still require human sign-off. The AI sends a notification, you confirm, it goes through. But Visa is also redesigning how it captures transaction data and handles disputes. They're trying to catch problems before they happen.
What's the endgame here?
Forestell basically said it: once you've approved the same transaction a thousand times, the agent asks if you want it to stop asking. Eventually, if this works, you'd set parameters and the AI just handles it. But that's years away. First, people have to trust it enough to let it happen at all.
Is Mastercard doing something similar?
They're in the space, but differently. Mastercard is focused on business AI agents buying services for companies. Visa is going after consumer shopping across its entire network. That's much bigger in scope.