What Broke OpenAI’s Checkout and What Gap Built Instead

Gap announced Today that shoppers will soon be able to buy products from its house of brands directly within Google’s Gemini. CNBC, which broke the story, reported it as a first for a major fashion company. Gap CTO Sven Gerjets told CNBC the service would launch “imminently.” The Gap announcement is best understood alongside the OpenAI one that preceded it, stalled, and was quietly restructured, because the distance between those two outcomes is where the operational requirements for agentic commerce actually live.

What OpenAI’s Retreat Revealed

OpenAI launched Instant Checkout in September 2025, powered by the Agentic Commerce Protocol built with Stripe, with Etsy, Walmart, and Shopify merchants among the first participants. OpenAI described the feature as the “next step in agentic commerce, where ChatGPT doesn’t just help you find what to buy, it also helps you buy it.”

The rollout did not hold. OpenAI could scrape some retailers’ websites to obtain product data, but that meant information about whether items were in stock, estimated delivery timing, or shipping costs could be inaccurate or out of date. “Crawling and scraping is inadequate to get the full breadth of product data that you need to do a good job of commerce,” one analyst told CNBC. Out of Shopify’s millions of merchants, roughly a dozen had gone live with ChatGPT checkouts before OpenAI shifted direction. Shopify president Harley Finkelstein confirmed the bottleneck was on the AI platform’s side, not the merchants’, and OpenAI had not yet built systems to collect and remit state sales taxes, a signal that transaction volumes never reached meaningful scale.

OpenAI and its retail partners headed back to the drawing board, with the company moving away from Instant Checkout and working with retailers to create dedicated apps within ChatGPT that reroute users to the retailer’s own website to complete a purchase.

How Google Built a Different Answer

Google announced the Universal Commerce Protocol at the National Retail Federation conference in January as an open standard covering the full shopping journey from discovery through post-purchase support, co-developed with Shopify, Etsy, Wayfair, Target, and Walmart, and endorsed by more than 20 ecosystem participants including Adyen, American Express, Best Buy, Macy’s, Mastercard, Stripe, The Home Depot, Visa, and Zalando.

The Gap arrangement is built on UCP. Gerjets told CNBC that the two protocols reflect different design premises: UCP was built for merchants to control the shopping experience, while ACP was designed more for discovery. Forrester analyst Chuck Gahun, writing in January, framed the operational distinction in specific terms: Google focuses on broader commerce journeys by providing contextual offers based on consumer searches, allows AI agents to authenticate customers and retrieve loyalty information, and permits agents to display discount data during checkout, while OpenAI has focused on compressing the commerce journey from discovery to instant checkout.

Under the Gap implementation, product information surfaced to Gemini shoppers is provided by Gap directly rather than crawled from its website, preserving data accuracy and keeping customer data within Gap’s systems. Google has updated its platform to load real-time product data, preventing problems such as out-of-stocks and pricing errors, and supports adding multiple items to carts and connecting loyalty memberships, two capabilities OpenAI has not yet fully addressed.

That architecture is a direct response to the data quality failure that grounded Instant Checkout. Andre Bechtold, president for SAP Industries and Experience, told attendees at NRF in January that simply “bolting on” AI tools to existing systems is not enough, and that without a strong data foundation, brands will be at risk because if customers get poor recommendations and errors in pricing, trust can disappear fast.

What This Requires Brands to Actually Build

Both protocols require structured product data fed directly by the merchant, and the distance between what most brands currently maintain and what either protocol needs to function is where most of the real implementation work sits.

Google has expanded Merchant Center to include new data attributes designed for conversational commerce, going beyond traditional keywords to include answers to common product questions, compatible accessories or substitutes, and what shoppers typically buy when something is out of stock. A product entry optimized for paid search carries title, price, availability, and an image. It does not carry the contextual attributes an AI needs to respond to a query like “what should I wear to a job interview” and complete a transaction on that basis.

Delivery terms are no longer only a post-purchase detail in agentic commerce. An AI shopping agent needs delivery windows, shipping costs, and returns terms that are structured and comparable across channels. If this information is missing or inconsistent, the agent is more likely to default to an offer that is easier to execute.

The implications differ by company type. For large retailers and top-tier brands with dedicated product information management infrastructure, the challenge is primarily one of prioritization: which attributes need enrichment, against which protocol’s schema, in which sequence. For mid-market brands and DTC companies scaling into wholesale, the catalog work is more foundational. Feeds built for a single retail relationship or tuned for keyword search are unlikely to be structurally ready for agentic surfaces without deliberate re-architecture. Coresight Research associate director John Harmon, speaking to the U.S. Chamber of Commerce in February, framed the competitive consequence directly: “near-term advantage will likely go to merchants whose catalogs are easiest for AI to interpret in natural language.”

The Loyalty Limitation and What It Points To

Gerjets acknowledged to CNBC that loyalty account linking and point redemption are not yet available through the Gemini checkout experience, though he said the capability is on the roadmap. The loyalty gap is a data problem by another name: the integration that would pass membership credentials, points balances, and tier status into an agentic transaction requires the same kind of structured, real-time data architecture that the catalog and delivery requirements do. For regular Gap, Old Navy, or Banana Republic customers whose purchasing behavior is organized around program membership, its absence creates friction at the point where the new channel most needs to convert.

A recent study found that only 13% of consumers report having completed a purchase after being referred by an AI assistant, while 70% said they are at least somewhat comfortable with an AI agent making purchases on their behalf. The distance between those two figures reflects the accumulated effect of unfamiliar checkout flows, unresolved questions about data handling, and missing integrations with existing customer relationships. The gap between comfort and completion is not specific to Gap’s rollout. The same structural friction appears wherever brands have moved faster on partnership announcements than on the data and integration work those partnerships require.

Where Both Protocols Go From Here

OpenAI and Stripe are continuing to develop the Agentic Commerce Protocol for app-based purchases, with the scope now focused on large integrated retailers building dedicated ChatGPT apps. Target, Instacart, and DoorDash are among those operating under the revised model.

Gerjets told CNBC that Gap intends to work with both platforms, framing the question of which protocol prevails as genuinely open. Forrester analyst Chuck Gahun’s January assessment recommended that brands and retailers experiment with both ACP and UCP in 2026 before sponsored content and ads arrive alongside chat responses, framing early participation as a way to establish presence before the economics of each platform shift.

Gap reported that online sales grew 4% in fiscal 2025 and represented 39% of total net sales, making digital the company’s fastest-growing revenue line. The Gemini partnership extends that trajectory into a surface where competitive position is still forming. eMarketer projects AI platforms will account for $20.9 billion in U.S. retail e-commerce sales in 2026, nearly quadruple 2025 figures. By 2029, eMarketer estimates that figure will surpass $144 billion, and the catalog, delivery, and loyalty integration decisions brands are navigating now are what will determine their position when that volume materializes.

Conversations On Retail

Conversations On Retail is a gathering place and resource center for retail and CPG executives, built to make it easier to stay current, discover the technologies and solutions shaping the industry, and connect with the people driving it forward.

We publish news, views, and reviews from staff editors, contributing experts, and trusted partners. Some articles are developed internally, while others are submitted by industry contributors or adapted from interviews and recorded conversations with industry leaders.