OpenAI Scales Back ChatGPT Checkout: Why Agentic Commerce Needs Universal Checkout Infrastructure
Arjun Bhargava
Co-founder and CEO @ Rye
Mar 5, 2026
8 minutes read
OpenAI scaled back ChatGPT checkout after 12 merchants went live. Here's why agentic commerce needs universal checkout infrastructure to work at scale.
TL;DR / Key Takeaways
OpenAI is scaling back native checkout inside ChatGPT, shifting purchases to retailer apps (Instacart, Target, Expedia) after only ~12 of Shopify's millions of merchants went live with the feature.
Shopify president Harley Finkelstein confirmed the bottleneck is on the AI firms' side — merchants are willing to participate, but the protocol-based onboarding process hasn't scaled.
The three structural blockers — merchant adoption lag, stale product data, and fraud safeguards for automated transactions — are inherent to any approach that requires merchant opt-in.
OpenAI and Stripe will continue developing the Agentic Commerce Protocol (ACP) for app-based purchases, but the scope is now narrower: a small pool of large, integrated retailers rather than millions of merchants.
Universal checkout infrastructure — which works on any merchant site via a product URL, no integration required — solves the exact problems that caused this pivot. Rye's Universal Checkout API already covers 15,000+ merchants with 99.9% order reliability and sub-5-second checkout on Shopify and Amazon.
Why Did OpenAI Pull Back on AI Checkout Automation in ChatGPT?
Six months after announcing checkout inside ChatGPT as a major business opportunity — partnering with Shopify, Etsy, and Stripe — OpenAI is pivoting away from native product-listing checkouts. An OpenAI spokesperson confirmed the shift: "Instant Checkout is moving to Apps, where purchases can happen more seamlessly." In practice, that means purchases will happen through individual retailer apps that plug into ChatGPT — like Instacart, Target, Expedia, and Booking.com — rather than directly from product listings in ChatGPT search results.
The numbers tell the story: out of Shopify's millions of merchants, roughly a dozen had actually gone live with ChatGPT checkouts. Shopify president Harley Finkelstein confirmed at an investor conference that the bottleneck isn't on the merchant side — it's the AI firms that haven't opened the doors wide enough.
OpenAI found that while users were researching products in ChatGPT, they weren't completing purchases there. The company also hadn't set up systems to collect and remit state sales taxes as of February — a clear signal that volume never materialized.
The Three Problems That Blocked Agentic Commerce Checkout in ChatGPT
This pivot highlights the core problems blocking AI-driven commerce at scale. These aren't OpenAI-specific failures — they're structural challenges that any protocol-dependent approach will face.
1. The Merchant Adoption Bottleneck
The Agentic Commerce Protocol (ACP) and Universal Commerce Protocol (UCP) both require merchants to explicitly opt in, integrate their catalogs, and maintain their product data for AI surfaces. OpenAI tried to shortcut this by partnering with Shopify and Etsy as aggregators, but even with those platforms handling the heavy lifting, adoption stalled at a handful of merchants.
This isn't a temporary problem. Implementing ACP or UCP requires merchants to build and maintain structured product feeds, integrate with new payment and catalog APIs, and keep pricing, inventory, and shipping data synchronized in real time across an additional channel. Enterprise merchants with dedicated engineering teams can absorb that lift — but it takes months. Small and mid-market merchants, who make up the overwhelming majority of online stores, often lack the engineering resources to justify the investment, especially when agent-driven transaction volumes are still negligible.
2. Product Data Standardization
Merchant product information — pricing, availability, shipping options — needs to be standardized and constantly updated for chatbots to access accurate data. This is the catalog problem, and it's one of the hardest problems in ecommerce. Static catalog feeds go stale. Prices change. Items go out of stock. Shipping costs vary by address.
Any system that relies on pre-ingested catalog data will always be working with information that's at least partially wrong.
3. Fraud and Trust Safeguards
Merchants and payment firms need safeguards against AI initiating fraudulent or erroneous transactions. Protocol-based systems address this through explicit merchant consent, but that leaves the non-opted-in merchants — the vast majority — completely out of reach for AI agents.
What This Means for Checkout API Infrastructure
Rye's Universal Checkout API was built specifically for the world that this OpenAI pivot now confirms. Here's why each of the three problems above has a different answer when you don't depend on merchant protocols.
No Merchant Integration Required
Rye works on any merchant site via a product URL. No opt-in, no protocol adoption, no catalog feed. The system uses agentic browser automation to navigate live checkout flows, which means coverage isn't gated by merchant participation. While OpenAI was waiting for merchants to come to them, Rye already works across 15,000+ merchants — including the vast long tail of stores that lack the engineering resources to implement ACP or UCP.
Real-Time Data, Not Stale Catalogs
Because Rye drives actual checkout flows with a real shipping address, it surfaces accurate landed cost — including real-time shipping and tax — at the moment of purchase. Sub-35-second offer resolution across any store. This solves the product data standardization problem by bypassing it entirely. You don't need a standardized catalog if you're reading the live checkout page.
The recently launched Product Data API extends this further, giving agent builders a discovery layer on top of universal checkout.
Built-In Fraud Mitigation
Rye's Merchant Risk Adapter uses residential IPs, geoproximity matching, and human-like interaction profiles to ensure automated orders aren't flagged and cancelled by merchant fraud systems. This is essential infrastructure for universal coverage — and it's a problem that protocol-based players don't solve because they assume merchants opted in natively.
The New Agentic Commerce Landscape
OpenAI's pivot clarifies the market into two distinct layers:
Protocol-based checkout (ACP/UCP) now serves a narrower use case: large, integrated retailers who build dedicated ChatGPT, Gemini, or Copilot apps. Think Instacart, Target, Expedia. These are high-value but few in number.
Universal checkout infrastructure covers everything else: the millions of merchants who won't build an app, DTC brands on custom platforms, legacy retailers, and the entire long tail of ecommerce. This is where the overwhelming majority of products live.
These aren't competing approaches — they're complementary layers. But this pivot makes clear that universal coverage can't be achieved through protocols alone. If you're building an AI commerce agent, you need infrastructure that works with the web as it exists today, not as protocols hope it will look in the future.
The Amazon Factor
One more detail worth noting: Amazon is investing $15 billion in OpenAI but isn't selling within ChatGPT. Amazon actually blocked AI apps from showing its products last year. Meanwhile, Rye already has direct Amazon integration with sub-5-second checkout. For agent builders, this means universal checkout infrastructure can offer access to the world's largest product catalog that even ACP can't provide.
Frequently Asked Questions
Is OpenAI shutting down shopping in ChatGPT?
No. OpenAI is shifting from native product-listing checkout to app-based checkout. Users will still be able to purchase through retailer apps plugged into ChatGPT (Instacart, Target, etc.), but the broad vision of buying directly from any merchant inside ChatGPT search results has been scaled back significantly.
What is the Agentic Commerce Protocol (ACP)?
ACP is a set of rules co-developed by OpenAI and Stripe that standardize how merchants, payments firms, and AI agents interact during transactions. It requires merchants to opt in and provide structured product data. OpenAI and Stripe will continue developing ACP for app-based purchases. For a full overview, see our guide to agentic commerce.
Why did only 12 Shopify merchants go live with ChatGPT checkout?
Shopify president Harley Finkelstein indicated the bottleneck was on the AI firms' side, not the merchants'. Getting merchants onboarded required hands-on work from OpenAI, and the company hadn't yet built foundational infrastructure like sales tax collection — suggesting volumes never reached meaningful scale.
What is universal checkout infrastructure?
Universal checkout infrastructure enables AI agents to complete purchases on any merchant website using just a product URL — without requiring the merchant to adopt a protocol, provide a data feed, or integrate with the AI platform. Rye's Universal Checkout API is the leading example, supporting 15,000+ merchants with 99.9% order reliability.
How does universal checkout work without merchant integration?
Instead of relying on merchant-provided data feeds, universal checkout uses agentic browser automation to navigate live checkout flows, extract real-time pricing and shipping data, and submit orders through the merchant's existing checkout — just like a human shopper would. Rye adds a caching layer that converts successful AI-driven flows into deterministic workflows for speed (sub-5s on Shopify and Amazon), plus a fraud-mitigation proxy so orders aren't flagged by merchant systems.
What Comes Next
The agentic commerce market is still early, and OpenAI's retreat from native checkout doesn't mean AI shopping is dead — it means the first approach didn't work at scale. The next wave will be built on infrastructure that meets the market where it is: merchants aren't going to adopt new protocols overnight, consumers aren't going to trust new payment flows immediately, and AI agents need to work with the web as it exists today.
That's exactly what Rye was built to do.
Product
Resources