
ChatGPT Wallet: OpenAI agents can now pay for online purchases
OpenAI has opened payment to its AI agents through a wallet built with Visa, Mastercard and American Express, currently limited to US Pro subscribers. The move shifts what AI agents can complete on their own, with direct consequences for checkout conversion and how brands reach buyers.
Blog du Modérateur reports that OpenAI now lets its AI agents make online purchases through a wallet developed with Visa, Mastercard and American Express. The feature is limited, for the moment, to Pro subscribers based in the United States. Beyond those three constraints, the report does not specify which merchants or product categories are covered, nor how merchants enrol.
What actually changes
Until now, an AI assistant could research a product, compare options and draft a cart, but a human still had to open a checkout and enter card details. Payment was the final barrier that kept the agent inside the conversation. Removing it means the handover to a site can happen after the purchase decision, or not happen at all.
The announcement does not say whether merchants keep a standard checkout flow, receive an order through a new channel, or how disputes, refunds and fraud checks are handled. Those gaps matter: any team tempted to build on this should treat it as a pilot until the payment rails and merchant terms are documented.
Consequences for acquisition and conversion
Search and social traffic has long been measured by clicks and sessions. If a share of buyers delegate purchasing to an agent, some revenue arrives with no session, no ad click and no browsing trail. That does not erase the previous funnel, but it changes what a dashboard can prove.
The practical questions are narrow. Does your product data survive being read by a machine rather than a person? Are prices, stock, delivery windows and return conditions stated plainly enough to be parsed? Is your checkout reachable and stable for automated sessions, or does it break under them?
There is also a monetisation angle. An agent that selects on the buyer's behalf rewards clear specifications, stable pricing and availability more than it rewards a persuasive landing page. Brands competing on promotional theatre may find that leverage shrinking.
What a team can do this week
- Audit your product pages as plain information: exact dimensions, materials, stock, shipping time, return policy, no ambiguity hidden in images.
- Check whether your analytics can separate human sessions from automated ones, so future anomalies do not look like a broken campaign.
- List the questions a buyer would ask before committing, and confirm each answer exists in text on the page.
The announcement leaves three things open: the full list of supported merchants, the fees or commercial terms, and any roadmap beyond US Pro subscribers. Teams should verify those with their payment providers before reallocating budget. The signal is clear enough, though: agentic commerce has moved from demos to a live payment rail, and product information is becoming a distribution asset. Eleutheros follows these shifts closely, from traffic sources to checkout economics.
Also worth reading
-
LinkedIn shares tips on getting cited by AI models
LinkedIn has published guidance on how companies can improve their chances of being cited by large language models. The report points to credibility and the authority of influencers as key factors, though the underlying mechanics remain undisclosed.
-
AI at work: study of 9,000 employees finds no clear time savings
An ActivTrak study of nearly 9,000 employees over six months challenges the idea that AI tools automatically save working time. For marketing and growth teams, it raises a practical question about where AI actually moves the needle.
-
More AI tools, less growth: what CMOs are getting wrong
A Digiday piece reports that enterprise CMOs see no growth from added AI spend, only task-level efficiency. The gap comes from siloed pilots and delegated ownership rather than from the models themselves.


