OpenAI drops ChatGPT checkout to pursue agent vision

OpenAI has scaled back work on a native checkout function inside ChatGPT, stepping away from plans that would have let users complete purchases without leaving the chat.

Internal teams had explored a deeper commerce integration with Shopify that would have connected product discovery and payment within ChatGPT. That work is no longer a priority, according to people familiar with the project.

The shift resets OpenAI's retail strategy. Instead of intermediating transactions, it is signalling a preference for partnerships and external commerce paths that keep payment, fulfilment, and post-sale processes with established retailers and merchants.

Checkout plans

The earlier concept aimed to make ChatGPT a shopping destination. Users would have been able to search for products, compare options, and buy items within the chat experience.

Known internally as ChatGPT Checkout, the initiative focused on reducing the steps between a user query and a completed purchase.

OpenAI hired eCommerce specialists for the project, including former Shopify staff, to add experience with retail systems and merchant tooling. Teams built prototypes designed to connect ChatGPT queries with merchant inventory data.

Strategic refocus

The native checkout tool was deprioritised after an internal review. Leadership concluded that building a full retail stack was not the best use of engineering resources, according to accounts of the decision.

OpenAI has also reduced resources directed at the Shopify integration, suggesting it does not want to anchor its commerce strategy to a single platform partner.

The refocus returns attention to core model development. It also reshapes how OpenAI might generate revenue from consumer products as it looks beyond subscriptions.

Payments and logistics

A native checkout system would have required OpenAI to handle sensitive payment information across many merchants. It also would have pulled the company into delivery, returns, and dispute handling.

That combination brings regulatory and compliance burdens in many markets and raises expectations around fraud controls, customer support, and risk management that resemble payment processor responsibilities.

Practical barriers added to the challenge. Shopify provides tools for merchants but does not standardise the full supply chain. A deep integration would have required adaptations for different merchant setups, shipping policies, and return terms.

Amazon speculation

Attention has shifted to the possibility of a partnership with Amazon. Such a relationship would align with an affiliate or referral model rather than an end-to-end checkout run by OpenAI.

Under that approach, ChatGPT could recommend products and send users to Amazon product pages. Amazon's large catalogue and established fulfilment network would reduce the need for OpenAI to build marketplace infrastructure.

Referral fees are a familiar model in consumer internet services. They can scale across countries without requiring a company to operate local payments and fulfilment systems, while shifting key operational responsibilities to retailers that already manage them.

Agent direction

The change also ties to a separate effort known as Operator, described as a computer-using agent that can carry out tasks by navigating websites and interacting with software.

This points to a different vision for shopping through ChatGPT: an agent that uses existing retailer sites rather than a dedicated buy button controlled by OpenAI. It would also reduce reliance on any single commerce platform.

That would move ChatGPT from answering questions to taking actions across the open web. In commerce, those actions could include searching for products, comparing offers, and completing checkout flows on third-party sites where users already transact.

Revenue pressure

OpenAI faces high costs to train and run large language models. It expects to reach annual revenue of USD $10 billion in the near future, according to information shared about its targets.

Subscriptions are a predictable revenue source, but commerce is often viewed as a large adjacent market for high-engagement consumer platforms. A referral-driven model typically carries lower operational risk than running payments directly.

Competitive context

The pullback changes the competitive picture for AI-driven shopping. Google and Microsoft have been integrating shopping features into their AI products, and both have advertising businesses and long-standing relationships with retail and commerce partners.

Moving away from a Shopify-linked checkout reduces the likelihood of a head-on push into merchant operations. It also leaves room for merchants and retailers to retain ownership of transactions, which can matter for customer data, brand experience, and support.

Consumer behaviour remains a key unknown. Many shoppers still prefer to complete purchases on familiar retail sites, especially for higher-value or more complex transactions where trust and a sense of control matter more.

An OpenAI spokesperson said:

"We are constantly evaluating how our models can be most useful to people in their daily lives, which includes exploring various ways to facilitate discovery and action."

OpenAI has reorganised teams connected to commerce work, with some staff moving to other areas and others leaving. The company is hiring more researchers and engineers focused on agentic behaviour, reflecting a longer-term bet that software agents-rather than a proprietary checkout layer-will shape how ChatGPT interacts with online retail.