OpenAI is stepping away from its most Amazon-like ambition for ChatGPT, cooling plans to turn the chatbot into a native e-commerce portal and refocusing on product discovery that routes shoppers to merchants’ own checkouts. The shift underscores a stubborn reality for AI assistants trying to do commerce at scale: closing the loop is harder than recommending what to buy.
Instant Checkout Stalls as OpenAI Refocuses Strategy
After testing an in-chat “Instant Checkout” that let users add items to a cart and pay without leaving ChatGPT, OpenAI is deprioritizing the feature. Company representatives have indicated the priority now is helping people find products, compare options, and then click out to merchants’ sites to complete purchases—rather than keeping checkout inside the chatbot.
Reports from The Information and CNBC describe a revised plan centered on merchant-built apps inside ChatGPT that guide shoppers to the seller’s own checkout flow. Behind the scenes, OpenAI is leaning on its Agentic Commerce Protocol, an open standard developed with Stripe, to normalize product data and support richer comparisons without owning the transaction itself.
Why AI Marketplace Dreams Are Hard to Execute
Turning a conversational agent into a marketplace is not just a UX challenge; it’s an operational gauntlet. Amazon’s marketplace works because it controls a dense stack—search, ads, Prime logistics, returns, and seller management—closing the loop with trust and speed. Third-party sellers now drive a majority of Amazon’s units sold, and its retail ad business brings in tens of billions annually, reinforcing that flywheel.
ChatGPT lacks that stack. Without Prime-like logistics guarantees, clear ownership of returns, and the downstream customer service guardrails buyers expect, an in-chat checkout can feel like an orphaned step. Merchants, meanwhile, are protective of first-party data, attribution, and upsell opportunities found on their own domains. Offloading checkout to a chatbot can weaken lifetime value and muddle marketing analytics—big reasons sellers favored links back to their sites over embedded payments.
There’s also the trust problem. Payment inside an AI interface asks consumers to share card data and shipping details in a new context. Even with robust security, behavior change is a moat. Industry benchmarks show that typical online retail conversion rates live in the low single digits; introducing a new, unfamiliar checkout tends to depress them further, not lift them.
Discovery Over Transactions in ChatGPT Commerce
OpenAI’s pivot puts discovery front and center. Using merchant feeds through its Agentic Commerce Protocol, ChatGPT will surface side-by-side visuals, price histories, feature breakdowns, availability, and review summaries. Think shopping comparison guide, not shopping cart. If the assistant can shorten research time and improve match quality, merchants still get warmed-up traffic—and keep their data and checkout systems intact.
That architecture mirrors what has worked elsewhere. Google leaned away from native “Buy on Google” and invested in its Shopping Graph to power comparisons that send users to retailers. Social platforms have also learned the hard way: Instagram dialed back the prominence of Shops after limited traction with native checkout, while TikTok Shop has grown but relied on aggressive subsidies and creator incentives to overcome trust and logistics gaps.
Early Signals From E-commerce Traffic and Ad Spend
OpenAI’s course correction tracks with usage data reported by third parties: studies last fall found that referral traffic from ChatGPT contributed little direct revenue to e-commerce sites, suggesting that users were comfortable asking for recommendations but hesitant to transact in-chat. Sources cited by The Information said users “weren’t using the chatbot to actually help them make purchases,” a gap that discovery-first design tries to bridge without fighting entrenched checkout habits.
For merchants, the economics also matter. If an assistant behaves more like a comparison engine, sellers can model it like upper-funnel search and measure it with familiar attribution. If it behaves like a marketplace, they must absorb new fees, cede data, and accept fragmented customer support—all for conversion rates that may not outperform their own sites. The math hasn’t penciled out yet.
What to Watch Next for OpenAI’s Shopping Strategy
Three indicators will show whether OpenAI’s new approach sticks.
- Data quality: Do merchants adopt the Agentic Commerce Protocol widely enough to enable accurate specs, real-time inventory, and authentic reviews?
- Monetization: Does OpenAI layer sponsored placements or affiliate-like economics, echoing how Amazon’s ad engine fuels its retail flywheel?
- Shopper trust: Do side-by-side comparisons reduce returns and increase post-click conversion on merchant sites?
OpenAI hasn’t abandoned commerce; it’s admitting that being Amazon inside a chat window is the wrong fight, at least for now. If the assistant becomes the best place to research a purchase and then hands off cleanly to a retailer, everyone has a clearer job to do—and the user gets fewer surprises at checkout.
