OpenAI drops plan for direct checkout inside ChatGPT

OpenAI is stepping back from its earlier plan to allow users to buy products directly inside ChatGPT, The Information reports. Instead, the company will route purchases through third-party apps connected to the chatbot.

According to the report, an OpenAI spokesperson said: “We are evolving our commerce strategy within ChatGPT to better meet merchants and users where they are. Instant checkout is transitioning to apps, where purchases can occur more seamlessly.”

This means that when users discover a product through ChatGPT search or recommendations, they will no longer complete the payment directly in the chatbot. Instead, they will be redirected to retailer apps or websites to finalize the transaction.

The update marks a change from OpenAI’s earlier ambitions in agentic commerce. In September last year, the company launched Instant Checkout inside ChatGPT. The idea was to allow users to discover products and complete purchases within the same interface.

OpenAI partnered with major commerce platforms, including Shopify and Etsy, bringing product listings from millions of online stores into the chatbot environment. The infrastructure behind the plan relied on the Agentic Commerce Protocol, which OpenAI developed with Stripe. The protocol was meant to create a shared framework for AI agents, merchants, and payment systems to handle product discovery and transactions.

User behavior revealed a key challenge

According to the report, one of the main issues behind the strategy shift appears to be user behavior. OpenAI found that people were using ChatGPT to explore products, compare options, and gather information. However, very few were completing purchases inside the chatbot.

Users seem comfortable asking ChatGPT questions such as which laptop to buy, where to find a cheaper flight, or which skincare product fits a certain skin type. The chatbot can respond with suggestions, comparisons, and links. But many users appear to prefer switching to a familiar retail environment where they already have accounts, saved payment methods, and order histories to make purchases.

At a recent investor conference, Harley Finkelstein, president of Shopify, said that only about a dozen merchants among the platform’s millions are currently using AI tools to sell products.

He said the slow expansion is tied to AI infrastructure still being developed. According to Finkelstein, the main constraint is that companies are waiting for AI agents and commerce integrations to mature.

Technical and compliance issues also slowed the rollout

The Information also reports that technical and regulatory challenges affected the rollout of direct checkout inside ChatGPT. 

Allowing a chatbot to complete purchases directly requires deep integration with merchant systems. The AI must pull live product data such as price, availability, shipping details, and inventory status from thousands or even millions of retailers at the same time. If any of that information is outdated, the transaction can fail or lead to incorrect orders.

Maintaining that level of synchronization is complex. Product catalogs and inventory systems must sync continuously with backend merchant databases to keep information accurate across large product catalogs. Some modern ecommerce platforms maintain synchronization intervals of just a few minutes to keep product availability accurate across hundreds of thousands of SKUs.

When a chatbot becomes the interface for shopping, that synchronization must happen instantly. If a user asks ChatGPT to buy a product that has just gone out of stock or whose price has changed, the system must detect that change before completing payment.

This challenge becomes more complicated when the platform is connected to millions of merchants. Retail systems often run on different commerce software, inventory tools, and pricing structures. Integrating those systems into a single checkout flow inside an AI assistant requires large-scale data standardization and constant updates across the entire merchant network.

Security and fraud prevention challenges

Security and fraud prevention also play a role according to The Information. Online payment fraud continues to grow, with global ecommerce losses estimated at $48 billion in 2023, according to Juniper Research. Because of this risk, ecommerce platforms run complex fraud detection systems that analyze thousands of signals for each transaction before approving a payment.

Traditional ecommerce checkouts already rely on fraud detection systems that examine location data, purchasing history, device behavior, and payment patterns. Introducing AI agents into that process adds another layer of complexity. Payment providers must determine how to verify that a purchase initiated by a chatbot is legitimate and authorized by the user. Without these safeguards, an AI system could trigger incorrect purchases or become a target for fraud attempts.

According to The Information, compliance requirements create another barrier. Large ecommerce platforms must comply with tax regulations, payment processing rules, and regional consumer protection laws. For example, companies operating in the U.S. are required to calculate, collect, and remit sales taxes across thousands of local tax jurisdictions. OpenAI had not yet built a system to collect and remit state sales taxes as of February this year. That capability is a standard component of established ecommerce platforms such as Amazon, Shopify, and eBay, which automatically handle tax calculation during checkout.

These technical and compliance requirements mean that turning ChatGPT into a fully functional checkout platform would require building much of the infrastructure that existing ecommerce companies have spent years developing. Without those systems in place, scaling direct purchases inside the chatbot becomes difficult.

OpenAI and Stripe continue work on AI commerce standards

Although direct checkout is being abandoned, OpenAI has not fully stepped away from building commerce infrastructure. Sources cited by The Information said OpenAI and Stripe will continue to develop the Agentic Commerce Protocol. The protocol aims to define how merchants, payment providers, and AI systems interact when purchases are initiated by AI assistants.

What OpenAI’s shift could mean for Google’s Universal Commerce Protocol

OpenAI’s decision to move purchases out of ChatGPT also brings attention to a similar infrastructure effort from Google. Google’s Universal Commerce Protocol is an open standard meant to support commerce interactions across different platforms and AI systems. The protocol is designed to allow product discovery, purchasing, and post-purchase services to work across multiple surfaces instead of being handled in a single interface.

That approach differs from the earlier concept OpenAI explored with direct checkout inside ChatGPT. Under UCP, the discovery stage and the transaction stage remain separate. AI systems can surface product information, but the actual purchase typically happens through the merchant’s own checkout system. That structure avoids requiring the AI platform itself to run payment processing, tax handling, and inventory management systems.

Those are the same areas that reportedly slowed OpenAI’s original checkout plans. Google’s UCP documentation says merchants provide structured product data that can be accessed by participating systems. That data can include product descriptions, pricing, availability, and fulfillment information. The idea is that different services, including AI assistants, can use that data during product discovery without controlling the transaction itself.

Google also already operates one of the largest product discovery systems on the web through Google Shopping and its merchant infrastructure connected to Google Merchant Center. Millions of retailers submit product feeds through Merchant Center so their listings can appear across Google surfaces such as Search, Shopping, and YouTube. Because those merchant feeds already contain structured data about pricing, product availability, and promotions, they provide a foundation that AI systems can use when helping users search for products. 

The approach also keeps the transaction layer with the retailer rather than the discovery platform. When users find products through Google services today, they typically complete the purchase on the retailer’s site or app. OpenAI’s new strategy for ChatGPT appears to move in a similar direction by relying on third-party apps for checkout.