NEWS · NOVEMBER 16, 2026 · AI

OpenAI introduces ChatGPT Wallet: dots agents will be able to pay without seeing your card details

On November 13, 2026, OpenAI announced ChatGPT Wallet. Its dots agents can pay for purchases without access to the real card details, and every purchase must be confirmed with a check like Face ID or a fingerprint scan. The rollout is limited and gradual, starting with ChatGPT Pro subscribers in the US.

01 · WHAT HAPPENED?

OpenAI gives its agents a payment layer

In an announcement dated November 13, 2026, OpenAI introduced ChatGPT Wallet. The company describes it as a payment layer that lets its personal agents, called dots, complete purchases for a user, and says it was built with American Express, Mastercard and Visa.

Per the announcement, the wallet reaches US ChatGPT Pro subscribers first, inside dots. OpenAI says it is starting small so it can learn from real use and improve compatibility with businesses, banks and payment networks. TechCrunch also reports that only a limited number of US Pro subscribers get access at the start.

02 · DETAILS

The card is added once, and every purchase is approved separately

In the flow OpenAI describes, a user adds a card once and then asks the agent to buy something. Before payment, the user sees the items, the total and the selected card, and confirms it with Face ID, a fingerprint or a similar check. Each approval leaves a record of what the user authorized.

After approval, the payment network supplies a payment token, a stand-in for the real card with its own number, security code and expiry date. OpenAI says the real card details are shared with neither the agent nor the business. TechCrunch adds that the agent passes the merchant what it needs to ship the order, such as an address and contact details.

The Verge reports the eligibility terms as: Pro subscribers who are at least 18, have a billing address in the US and use an eligible card issued in the US by American Express, Chase, Mastercard or Visa. OpenAI says the card keeps its usual rewards, and its dispute protections still apply in eligible cases such as a defective item. Merchants continue to control pricing, fulfillment, returns and customer service.

A fuller hand-off is planned but undated: OpenAI expects that users will later be able to pre-approve purchases within limits such as a spending cap, a category or a time window. Today every purchase asks for its own approval.

03 · WHY IT MATTERS

The hard part is no longer finding the product but completing payment safely

OpenAI says 1.2 billion people use ChatGPT each week and that many of them use it to decide what to buy. The wallet moves the agent from product research to payment. That figure is OpenAI's own claim, so treat it as unverified by outside data.

On the merchant side OpenAI is explicit: sites must not block agents by mistake, and payment and risk systems must recognize legitimate agent purchases. Its Commerce Partner Playbook, published alongside, points to keeping product data current, configuring CDNs and firewalls to recognize signed agent requests, and working with payment and risk providers. OpenAI names Forter, Riskified, Sardine and Signifyd as fraud-prevention partners and Adyen, Checkout.com and Stripe as payment service providers.

There is context on the other side too. The Verge notes that some retailers already block agents, and TechCrunch says Amazon is the largest retailer to do so, steering users toward its own agent.

04 · TÜRKİYE

What it means for businesses in Türkiye

This section is our reading, not a fact from the sources. Per the sources, the first rollout is limited to US Pro subscribers and US-issued cards, so we assume users in Türkiye are outside it for now.

The signal for e-commerce businesses is still clear: the technical requirements of agent-led shopping are already being written down. A store that sells to US customers, or plans to, can check that its bot rules and firewall do not block legitimate agents, that product data is current, and that its risk system does not reject agent purchases without cause. That is our suggestion.

The UNALSOFT view

Agents that can pay raise the question of approval, authority and record-keeping in any automation project. We read this news as one example of a wider question in agent-based workflows: which step needs a human sign-off, and which data should never reach the agent at all. In our agentic AI work we settle those boundaries at the start of workflow design. The news gives no Türkiye date, so this is a development to plan around, not a step that needs a decision today.

Which step in your agent workflows should require human approval?

A short conversation is enough to define your processes and authority limits together.

Message on WhatsApp