This spring, OpenAI quietly discontinued its instant checkout feature, deciding against allowing users to make payments directly within the chat interface. Instead, they established a clear division of roles: ChatGPT handles product discovery and comparison, while the actual checkout process redirects users to the merchant’s website or a third-party app. Reports from Modern Retail and CNBC have confirmed this shift. Both merchants and platforms agree that while consumers enjoy turning to AI for recommendations and product research, they are reluctant to click “Buy” inside the chat window. Coupled with limited product selection and slow data updates, in-chat payments never gained widespread adoption.
This might sound like the mere elimination of a simple feature, but dismissing it as gossip within the tech world would mean missing the heart of the matter. It reveals a major shift in the AI business landscape: the bifurcation of the shopping process, where gaining exposure is one thing, and ultimately closing the sale is quite another.
As the saying goes, you cannot have your cake and eat it too. Over the past six months, the market attempted to bundle “product discovery” with “checkout,” hoping that AI agents could handle the entire journey from product selection to placing an order. Now, the dust has settled: product discovery takes place within the conversation, while the actual transaction reverts to the merchant’s storefront. For merchants, this is neither a victory nor a defeat; it simply means the boundaries of their business have become more clearly defined. While merchants retain control over core elements—such as pricing, inventory, customer loyalty, and logistics—they face a tougher challenge: when a buyer’s purchase intent first emerges in a chat window, what specific qualities will land your product on the AI’s “shortlist”? Not to mention the need to handle thorny real-world issues like product returns.
Meanwhile, the tech sector has diverged regarding technical approaches. On one hand, OpenAI and Stripe are championing the “Agentic Commerce Protocol” (ACP). This protocol mandates the use of structured product data streams to ensure accurate product indexing and to move beyond a reliance on in-chat checkout; retail giants such as Target, Sephora, and Nordstrom have already joined this camp. On the other hand, Google and Shopify support the “Universal Commerce Protocol” (UCP). Under this model, merchants simply list their service capabilities, while the AI identifies specific products and guides the entire shopping journey, including checkout. Developers are even building bridges for interoperability so that a single shopping AI can support both protocols. Additionally, Visa and Mastercard are exploring methods for secure AI-driven payments. Despite the variety of technical paths, they all point to a single core requirement: machines must understand the products being sold.
However, the proliferation of these agreements has not made life easier for merchants; they merely establish a baseline. Your data sources must be structured, product attributes standardized, and pricing and inventory information accurate in real-time. Industry blogs are increasingly championing AEO and GEO (Generative Engine Optimization)—optimizing specifically for AI-driven Q&A engines. The competitive focus is shifting: the priority is no longer just click-through rates, but “Share of Model”—essentially, the “voice” or prominence your brand commands within AI-generated answers. Yet, the failure of “Instant Checkout” features serves as a cautionary tale: successfully entering the arena is only the beginning; when AI actually recommends your product to users, will it stand up to scrutiny?
Data from IBM, McKinsey, and Adobe point the same way. More people are letting AI handle parts of their shopping. AI-driven retail traffic is multiplying fast. The numbers will change, but the structure won’t. Buyers start with the AI. Your website is no longer the only front door. It’s a landing page you have to earn. If the AI recommends the wrong size, an out-of-stock item, or an expired sale, trust breaks down. It hurts the AI, and it hurts your brand. If the discovery phase fails, a slick checkout page won’t save you. You’re just running fast in the wrong direction.
Consequently, the core issue has shifted. The focus is no longer on whether to venture into AI-driven shopping, but on a practical, day-to-day question: Is your product information consistent across all channels? Do your reviews, specifications, and policies back up your marketing claims? If a sale is lost, what is the reason? Is it a lack of visibility? Insufficient data? Or did a competitor offer more compelling evidence? An exposure score acts like a thermometer: it tells you if you have a fever but cannot reveal the underlying cause. The true “root cause” lies within the conversion funnel. Having an “AI-friendly” page is one thing; actually being “chosen by AI” and recommended to users is quite another.
Think of it this way: the protocol is like the lane markings on a highway, while your product data feed serves as the traffic report. However, whether the GPS ultimately guides the driver to your store is entirely up to you. Is your store reliable? Is it open for business? Are your prices reasonable? Even though the road standards are uniform, the quality of the destination is determined solely by the merchant.
As the philosopher John Locke observed, without clear rules, success cannot be fairly measured. Today, these rules manifest as transparent AI decision-making and traceable operational actions. While models can offer recommendations and platforms facilitate data flow, merchants require something more—a layer of governance and intelligence that sits above these underlying protocols. They need their products to be understood, compared fairly, and given a genuine opportunity to be recommended. This lies at the heart of the Hasmord Merchant OS philosophy: ensuring that every product is truly prepared for the context of AI-assisted environments.
Whether the checkout process is handed back to retailers or remains within the AI chat interface, it does not signal the conclusion or uncertainty of the AI shopping era. While the initiative in product discovery has shifted, the power to actually sell those products remains as firm as ever. Today, the wise are quietly stocking their digital shelves, while the noisy are still debating which protocols to adopt. Protocols will inevitably evolve, but what truly matters is making the “shortlist.” Is your name on that list? More importantly, do you have a compelling reason to be there?