ChatGPT Wallet: What It Is and What It Means for E-commerce

OpenAI launched ChatGPT Wallet on November 13, and the headlines arrived on cue: AI agents can now buy things for you.

That's technically true. I also think it's the least interesting way to read the news, and for e-commerce brands it's a misleading one.

Here's my take. ChatGPT Wallet isn't about AI deciding what you buy. It's about removing the single most expensive moment in AI-assisted shopping, when the customer has to leave the conversation to pay, and making the payment safer than the checkout it replaces. If you run an e-commerce business, that's a much bigger deal than robot shoppers, and it changes where your sale is actually won.

What actually shipped

First, the facts, because the hype has drifted away from them.

Wallet was built with American Express, Mastercard and Visa. You add a card once. When one of ChatGPT's agents (OpenAI calls them "dots") is ready to check out, you see the items, the total and the card, and you approve that purchase with Face ID or a fingerprint. Every purchase, every time.

Neither the agent nor the merchant ever sees your card number. The networks issue a single-transaction token with its own number, security code and expiry. Your card rewards and eligible dispute protections still apply.

On the merchant side, OpenAI says existing checkout pages should work without a Wallet-specific integration. The agent completes the purchase on the merchant's own site through a cloud browser, or through a commerce plugin if the merchant builds one. Merchants keep pricing, fulfilment, returns and the customer relationship, and OpenAI says it isn't taking a cut of transactions for now.

The rollout is deliberately small: US only, ChatGPT Pro subscribers aged 18 and over, eligible cards, inside dots. As Amex's Luke Gebb told CNBC, the idea is to "start it in a contained environment."

Wallet vs. Instant Checkout: move the wallet, not the store

OpenAI has tried this before, and the difference between the two attempts is the whole story.

Instant Checkout, launched in fall 2025, tried to move the store into ChatGPT. Merchants had to integrate through OpenAI's Agentic Commerce Protocol so ChatGPT could run the checkout itself. It was scaled back in March. Very few merchants went live, few users tried it, and brands didn't want to become anonymous fulfilment centers inside someone else's interface.

Wallet moves the wallet into ChatGPT instead. The merchant keeps its own store and its own checkout, with no new integration required. What ChatGPT owns is the payment credential and the moment of approval. The customer stays in the conversation, the agent goes to the merchant's site, and the card number never changes hands.

That's a much smarter design, and it's why I think this version sticks. The goal was never to turn ChatGPT into a marketplace. It's to let people finish a purchase without leaving the platform, and to make that purchase more secure than the one they'd have made on their own.

The friction was never the decision

I've spent a lot of time looking at how people actually shop through AI. The pattern is consistent: the conversation does a great job, and then the experience falls apart.

Someone asks ChatGPT for the best running shoes for flat feet under $150. They get three good options and a clear recommendation. They've effectively decided. Then they click out, land on a product page, hunt for their size, get asked to create an account, type a card number on a phone, re-enter their shipping address and maybe hit a CAPTCHA. Every one of those steps is a chance to lose them.

The buying decision was made in the chat. The friction was everything after it. Wallet removes that friction.

So when people frame this as "agents making purchases," I push back. The person decides, and the agent does the paperwork. The per-purchase biometric approval isn't a temporary measure on the way to fully autonomous shopping. For most commerce, it is the product. People want help researching and buying, and most of them don't want to hand over the decision.

Will agents eventually buy on their own? For some categories, yes. OpenAI has said it wants to support purchases within pre-set spending limits, categories and timeframes, and that will work well for replenishment: dog food, printer ink, the moisturizer you reorder every six weeks. But that's a narrow slice of commerce. Considered purchases will keep a human in the loop. What's changing is that the yes now happens in the same place as the research.

Security is the half nobody is talking about

The other part of this story is security, and it's being underrated.

Today, paying for something you discovered through AI means typing your card into whatever site the AI sent you to. Sometimes that's a well-run brand. Sometimes it's a store that went live last Tuesday. The customer has no good way to tell.

With Wallet, the card number never leaves the network. The agent gets a scoped, single-use token, so a leaked token isn't a leaked card. Chargeback rights stay intact. For the shopper, this is arguably safer than the checkout it replaces.

That matters a lot for independent brands. "I've never heard of this company, so I'll just buy it on Amazon" is one of the biggest hidden costs in DTC. A trusted wallet with tokenized payments and clear dispute rights lowers that barrier. If the customer trusts the wallet, they don't need to trust your checkout page as much. That's a real opening for challenger brands, and I expect the smart ones to take it.

We've seen this movie before. It's called Apple Pay.

If this pattern feels familiar, it should.

In October 2014, Apple launched Apple Pay. Apple didn't build stores or ask retailers to rebuild their checkout. It put your existing cards into the phone, replaced the card number with a network token, and made you approve each payment with your fingerprint. The merchant's terminal stayed the same. What changed was that you never had to take out a physical card again, and paying became both faster and safer.

Swap the phone for ChatGPT and the fingerprint for Face ID, and that's ChatGPT Wallet. Same tokenization, same biometric approval, same "bring your existing cards" model, same decision to work with the merchant's existing checkout instead of replacing it. Apple Pay is now used by hundreds of millions of people and accounts for roughly half of US in-store mobile wallet use, by most industry estimates.

Then Apple made its next move. Once it owned the moment of payment, it launched its own card.

Apple Card arrived in August 2019, issued by Goldman Sachs on the Mastercard network. It was designed around Apple Pay: Daily Cash paid more when you used the card through Apple Pay, and the most at Apple itself. By January 2024, Apple said it had more than 12 million cardholders, had paid out more than $1 billion in Daily Cash in 2023 alone, and that its linked savings account had passed $10 billion in deposits.

The banking side was messier. Goldman reportedly lost billions on its consumer push, with Apple Card a big part of it, and in January 2026 JPMorgan Chase agreed to take over the program, about $20 billion in balances, at a discount of more than $1 billion. Apple Pay Later was shut down in 2024. But look at who came out ahead. Apple kept the customer, the interface and the brand, and let the bank carry the balance sheet.

My hot take: a ChatGPT card is coming

ChatGPT Wallet is the Apple Pay step. I think the Apple Card step comes next.

My bet is that OpenAI launches a ChatGPT-branded credit card within the next six months, issued by a bank partner, and very likely with one of the networks or issuers it just launched Wallet with. The case is hard to ignore:

• The distribution is already there. ChatGPT has more than 1.2 billion weekly users and nearly 100 million paying subscribers. Apple Card launched into a much smaller pool of iPhone owners in the US and still passed 12 million cardholders.
• OpenAI needs the revenue, and it isn't taking a cut yet. OpenAI has said it isn't charging merchants on Wallet transactions for now. A co-branded card earns from every purchase without taxing merchants, which is a much easier story to tell the market ahead of a possible IPO.
• Rewards change behavior. Apple used Daily Cash to make Apple Pay the default. A ChatGPT card that pays extra on purchases made in ChatGPT would do the same thing: give people a financial reason to buy where the AI recommends.
• It closes the loop. OpenAI now has ads, business profiles and a wallet. A card connects the recommendation to the transaction, which is exactly the data an ads business needs.
• Apple already showed the template. Own the interface, own the customer, and let the bank take the credit risk.

If I'm right, the implication for brands is uncomfortable but clear. Once shoppers are rewarded for buying inside ChatGPT, the AI's shortlist won't just be where decisions get made. It'll be where the money flows by default. Which brings me to what I'd actually do about it.

What I'd do about it: win the shortlist

If discovery, comparison and payment all happen in one conversation, the sale is decided at the moment ChatGPT builds its shortlist. Everything that matters now happens before checkout, which means this is a visibility problem, and SEO and GEO are how you solve it. Here's what I'd prioritize.

1. Treat AI visibility as the top of your revenue funnel

When checkout lives in the chat, there's no second-chance visit to your site where better design or a retargeting ad wins the customer back. If ChatGPT doesn't put you in its shortlist, you don't lose a click. You lose the sale, and you never see that it happened.

That's why we built Passionfruit around AI visibility. You need to know which buying prompts you show up for, how assistants describe you against competitors, and which sources those answers rely on. If you aren't measuring this yet, start with how to monitor your brand across ChatGPT, Perplexity and AI search. A year ago, this was a forward-looking metric. With Wallet, it's a revenue metric.

2. Claim your business in ChatGPT this week

Look at the timing. Three days before Wallet, on November 10, OpenAI launched Businesses in ChatGPT. Brands can now claim and verify a ChatGPT profile, add product feeds with pricing and availability, see their first data on mentions and clicks from inside ChatGPT, and, for eligible US businesses, run a Business Agent that talks to customers in the chat.

Put the two launches together and you can see what OpenAI is building: a verified merchant profile, a brand agent that answers questions, and now a wallet that can pay. That's a storefront inside ChatGPT, from discovery to checkout. I'd bet these pieces get stitched together quickly.

Claiming your profile won't make ChatGPT recommend you. OpenAI is explicit that it doesn't guarantee placement. But it does decide whose version of your prices, policies and stock the model sees, and the insights it unlocks are the first visibility data you'll get from inside ChatGPT itself. If you haven't claimed yours, start there. We covered what launched and how to claim it earlier this week.

3. Earn the mention with GEO

A claimed profile tells ChatGPT the facts about you. It doesn't make ChatGPT pick you. That still comes down to the pages, reviews and third-party sources the model reads when it answers a buying question, and that's the job of generative engine optimization.

In practice, that means:

• Pages that answer the questions shoppers actually ask ("best X for Y under $Z"), not just product pages built for category keywords.
• Clear comparisons against the alternatives you're shortlisted with, so the model doesn't have to piece them together from competitors' content.
• Presence in the reviews, roundups and communities that AI engines cite for your category.
• An SEO foundation that's still solid. ChatGPT searches the open web, and pages that don't rank or get crawled rarely get cited.

This is earned, not bought. ChatGPT now has ads, and they have a place, but paid placement sits below the answer while GEO earns a spot inside it. When the answer is also where the purchase happens, being inside it is what counts.

4. Make your catalog easy for AI to read

Agents recommend what they can understand. That means clean, structured product data: accurate prices, live inventory, variants, shipping times and return policies, consistent across your site, your marketplace listings and your ChatGPT profile. Every fact you expose directly in a product feed or structured data is one the model doesn't have to guess from your HTML, and guessing is where brands get misrepresented or skipped. Our product feed guide for ChatGPT Shopping and AI commerce covers the fields most brands miss.

5. Make sure AI can actually reach you

This is technical SEO for the agent era. United Airlines has said it will block agents, and Amazon already does. Those are deliberate choices by companies with enormous direct traffic. Most mid-market brands that block AI do it by accident, through a bot rule or firewall setting nobody has looked at in years.

If ChatGPT's crawlers and cloud browser can't load your pages, you can't be read, recommended or bought from. OpenAI's guidance is to allow verified ChatGPT browser traffic, which identifies itself using Web Bot Auth. Check your robots.txt, your CDN's bot settings and your firewall rules before you spend another dollar on content. Content that AI can't reach doesn't count.

Where I think commerce is going

My bet is that checkout is moving to wherever the customer already is, and for a fast-growing share of shoppers, that's a conversation with an AI.

This week made that concrete. OpenAI shipped a business profile on Tuesday and a wallet on Friday. The pieces of a full storefront inside ChatGPT are now on the table.

That doesn't kill the brand website. It stops being the only place a purchase can happen and becomes one of several. Your site will still matter for brand, content and loyalty. More and more, though, the transaction itself will happen in an assistant, and the brands that win will be the ones agents can find, understand and buy from without friction.

I'd also push back on the dystopian version of agentic commerce, where autonomous agents drain bank accounts on a whim. Wallet is designed the opposite way: scoped, tokenized and approved by a human every time. The future I see is less about robots shopping for us and more about shopping finally working the way people always wanted it to. You ask, you decide, you confirm, and it's done.

Today, Wallet is small: US only, Pro only, inside dots. That's how these shifts always start. ChatGPT has more than 1.2 billion weekly users, and I don't expect payments to stay limited to Pro subscribers for long. The brands that build their AI visibility and clear out agent friction now will already be in the shortlist when it opens up.

Want to know whether ChatGPT recommends your products, and what it says about you when it does? Book a call with the Passionfruit team.

Sources: CNBC · OpenAI · OpenAI Developer Community · The Decoder · Forrester · Apple Newsroom · TechCrunch · 9to5Mac