ChatGPT Can Now Buy Things for You. Here's What Happens If a Purchase Goes Wrong.

ChatGPT Wallet lets AI agents complete purchases with your credit card. OpenAI's terms explain who handles incorrect orders, unauthorized charges, and Wallet errors, while leaving customers responsible for reviewing what they approve.

Written By
Marianne Sison
Marianne Sison
Nov 16, 2026
5 minute read

Your AI assistant finds the perfect pair of headphones, places the order, and charges your credit card. A few days later, the wrong model arrives.

Who fixes the mistake?

OpenAI introduced ChatGPT Wallet, where AI agents complete purchases on a user's behalf. The feature launched November 13, with built-in payment security and a requirement for customers to approve transactions.

But OpenAI's terms reveal something consumers should understand before handing their shopping lists to an AI agent: customers remain responsible for transactions they authorize, while the company directs most payment and order disputes to merchants or card issuers.

The distinction becomes particularly important when an AI agent misunderstands instructions or submits an incorrect order.

ChatGPT Wallet lets AI agents shop with your credit card

Advertisement

ChatGPT Wallet is gradually rolling out through dots, OpenAI's personal AI agents, to US ChatGPT Pro subscribers aged 18 and older. Users need an eligible US-issued credit card and a US billing address.

Here's how it works.

You ask your dot to find a product or make a reservation. The agent searches for options and prepares the checkout. Before payment, Wallet displays the merchant, items, total, and selected card.

You then approve the purchase through biometric verification, such as Face ID or Touch ID.

Once approved, the payment network generates a temporary substitute for your card details, known as a payment token. It contains the information necessary to process the payment, while your actual card number stays hidden from the AI agent and merchant.

OpenAI developed Wallet with American Express, Mastercard, and Visa. Customers also retain their existing card rewards and protections for qualifying disputes.

However, approval doesn't guarantee a successful purchase.

The card issuer must authorize the payment, and the merchant must accept the order. OpenAI acknowledges that some retailers may reject legitimate agent purchases because their fraud-detection systems don't recognize automated shoppers.

Advertisement

Users should therefore check for a merchant confirmation before assuming their order went through.

What happens when ChatGPT buys the wrong item?

The most revealing part of ChatGPT Wallet may be its purchase authorization terms.

Section 4 explicitly warns: "AI systems can make mistakes."

It tells customers to review their purchase details carefully before approval. By selecting Approve, users authorize OpenAI to submit the displayed order and associated payment information to the merchant.

Consider two possible situations.

First, an AI agent selects a $150 pair of headphones instead of the $250 model you requested. The approval screen identifies the cheaper headphones, but you approve the transaction anyway.

Under OpenAI's terms, you've authorized the purchase shown on the screen. You could request a return from the retailer, but the outcome would depend on the merchant's policies and any applicable cardholder protections.

Now consider a different mistake.

Your approval screen shows the correct $250 headphones, but the agent submits an order for the $150 model.

Advertisement

OpenAI says approval applies only to the specific purchase details presented to the customer. If a material detail changes before submission, Wallet must request another approval.

The company's support documentation directs customers to OpenAI Support when a purchase differs from what they approved.

However, it doesn't promise automatic reimbursement for every agent-generated mistake.

Who pays when something goes wrong?

OpenAI's terms distinguish between merchant problems, payment disputes, and Wallet errors.

For an incorrect or undelivered product, the merchant handles the complaint. Customers should contact the retailer about cancellations, returns, and refunds. OpenAI identifies the merchant as the seller responsible for fulfilling the order.

For an unauthorized charge or billing error, customers should contact their card issuer. Existing dispute rights still apply, including protections under applicable credit-card laws.

For a Wallet malfunction or a purchase that differs from the approved transaction, customers should contact OpenAI Support. The company may examine its purchase records or provide information relevant to the investigation.

Advertisement

Here's where OpenAI's legal terms get interesting.

Section 7 contains a broad disclaimer of liability for activities involving ChatGPT Wallet. The company says customers should seek resolution through their agreements with merchants, payment networks, or other relevant providers.

Meanwhile, Section 14 of OpenAI's Service Terms states that users are responsible for transactions made on their behalf by dots and related agentic features.

These provisions put considerable responsibility on customers, although OpenAI also acknowledges that its Wallet terms cannot eliminate rights protected under applicable law or cardholder agreements.

For US credit-card holders, the Consumer Financial Protection Bureau advises reporting billing errors promptly. To preserve certain legal protections, customers generally must send a written billing-error notice within 60 calendar days after the charge first appears on their statement.

A transaction authorized through an AI agent may still qualify for applicable dispute protections. The outcome depends on what happened, the relevant payment rules, and the customer's legal rights.

Advertisement

OpenAI keeps records of what customers approve

OpenAI has introduced another safeguard that could become important during disputes: purchase authorization records.

According to its Wallet Privacy Notice, the company collects information about the purchase request, the action a customer approved or rejected, and the authentication result.

It also records transaction information, including what was submitted to the merchant.

These records could help investigators determine whether an agent followed the customer's authorization.

For example, comparing the approved purchase with the merchant's order could reveal whether an error occurred before approval or during checkout.

However, a transaction record doesn't automatically establish who must reimburse the customer. The applicable agreement and consumer-protection laws still govern the dispute.

Advertisement

What happens when AI agents no longer need approval?

OpenAI eventually wants users to delegate purchases through predefined spending limits, permitted categories, or time periods.

Instead of approving every transaction, customers could authorize an agent to make purchases within those boundaries.

The company describes this as a future capability. At launch, individual purchases still require biometric approval.

Greater purchasing autonomy opens another question: What happens if an agent technically stays within a spending limit but buys something the customer never intended?

As The Neuron previously reported, financial institutions are already examining how to establish what customers have authorized AI agents to do.

For consumers, reviewing purchase details remains essential. They should also check merchant confirmations, monitor card statements, and report unfamiliar transactions promptly.

ChatGPT Wallet lets AI agents pay through existing card networks, and its authorization records provide evidence of what users approved.

Advertisement

The next challenge is determining how those records will be used when an agent makes a costly mistake, particularly once customers can authorize purchases in advance.

A successful checkout shows that an AI agent can spend your money. Resolving the next disputed transaction will help establish how much responsibility comes with that convenience.

Marianne Sison

Marianne is a technology analyst with nearly five years of experience reviewing collaborative work management solutions. She helps businesses identify the right tools and apply best practices to streamline workflows and improve project performance. Her insights on project management and unified communications appear in publications like Project-management.com, TechRepublic, and Fit Small Business.

The Neuron Logo

Don't fall behind on AI. Get the AI trends & tools you need to know. Join 700,000+ professionals from top companies like Microsoft, Apple, Salesforce and more.

Property of TechnologyAdvice. © 2026 TechnologyAdvice. All Rights Reserved

Advertiser Disclosure: Some of the products that appear on this site are from companies from which TechnologyAdvice receives compensation. This compensation may impact how and where products appear on this site including, for example, the order in which they appear. TechnologyAdvice does not include all companies or all types of products available in the marketplace.

Stay in the loop

Get notified when we publish new articles.