Twelve merchants.
Out of millions of Shopify sellers, roughly a dozen had gone live with native checkout inside ChatGPT when OpenAI pulled the plug. Shopify says the number was closer to 30. Either way, a rounding error on a rounding error.
I talked to insiders. The reason isn't complicated. The team at OpenAI leading commerce has no commerce experience. Their one commerce-related job posting doesn't include the word "commerce" anywhere in the qualifications. OpenAI calls this "the blessing of inexperience." That philosophy produced ChatGPT. It did not produce a working checkout.
Because commerce is not a reasoning problem. It is a plumbing problem. Payments, fraud, tax collection, inventory sync, returns, each one a decade-deep operational discipline. As of February, OpenAI hadn't built a system to collect state sales taxes. Product feeds were scraped without real-time inventory. No assortment plan for the channel. ChatGPT was surfacing out-of-stock items that were never meant to be sold there.
And now we have the clearest proof yet that the model was broken. Walmart just scrapped Instant Checkout entirely and replaced it with something far more interesting. Instead of letting OpenAI handle the transaction, Walmart is embedding its own Sparky shopping agent directly into ChatGPT and Gemini. When a user asks ChatGPT to find paper towels, the request routes to Sparky, which searches Walmart's inventory, presents options, and processes the order. Walmart keeps the customer data, the transaction, and the post-purchase relationship. OpenAI gets a flat API fee.
The early numbers are telling. Walmart's internal data shows users who access Sparky through ChatGPT complete purchases at roughly 70% the rate of Walmart.com, far better than anything Instant Checkout delivered. The difference is trust. Customers recognize they're interacting with Walmart's agent, even inside someone else's app.
This is the model. Not "buy inside the chatbot." The retailer owns the agent. The AI platform provides distribution. What failed was OpenAI owning the transaction. What is working at enormous scale is AI-driven discovery.
ChatGPT processes 50 million shopping queries a day. Researching products is the top use case among answer engine users. Completing a purchase inside the engine was dead last. Fifty million queries, almost nobody clicking "buy" until Walmart handed the buy button to its own agent.
Amazon has blocked 47 AI crawlers from its site. That means roughly 40% of U.S. e-commerce — your listings, your reviews, your products — is invisible to the fastest-growing product discovery channel on the internet. Walmart is embedding its agent into ChatGPT and Gemini. Amazon is blocking them. One retailer is capturing AI-driven demand. The other is pretending it doesn't exist.
If your entire catalog lives behind Amazon's wall, you are optimizing for a discovery channel that is losing share to one where you don't show up. Structured product data, off-Amazon presence, and content built to answer questions rather than stuff keywords are no longer nice-to-haves. They are the price of being visible where demand is increasingly being shaped.
The buy button didn't matter. Who owns it always did.
Lumian can help you get discovered on Rufus. Book a Free Consultation.
In this week’s issue:
Marketplace: Prime Day, Sponsored Prompts, Shop Direct
Seller Central Update of the Week: Product Opportunity Explorer
Tweet Spotlight: Seller Evolution
Marketplace Madness
Amazon is reportedly considering moving Prime Day from its usual July timing to June this year, although the company has not confirmed it yet. If that happens, the biggest midyear sales event on Amazon would land earlier on the calendar, and sellers would be dealing with Prime Day prep sooner than usual. Last year’s event expanded to four days, which gave shoppers more time to buy but reduced some of the urgency that usually helps drive fast conversion.
Why it matters:
Even before anything is official, this is a signal to tighten your timeline. If Prime Day shifts into June, sellers may need to move up inventory buys, deal planning, creative refreshes, and ad spend earlier than expected, while also preparing for competitors like Walmart and Target to launch overlapping promotions that make the whole period more expensive and noisy.
Lumian is your extended team of dedicated human experts + AI employees that help you scale much faster on Amazon using AI.
Amazon is rolling out a new ad feature called Sponsored Products prompts and Sponsored Brands prompts that adds AI-generated product explanations inside sponsored ads. The system uses information from your product listings, Brand Store, and campaign data to automatically surface helpful details to shoppers while they browse or consider a purchase. The feature has been in beta since late 2025 and becomes generally available in the U.S. on March 25, 2026. All existing Sponsored Products and Sponsored Brands campaigns will be automatically enrolled and any shopper interactions will be billed through the normal cost-per-click model.
Why it matters:
Your ads may start answering shopper questions directly inside the buying flow, which could improve conversion if your listings are clear and information-rich. The flip side is that weak listings or poorly structured product pages will now feed the AI that speaks for your product, so optimizing titles, bullets, images, and Brand Store content becomes even more important. Over time, this effectively turns your listing quality into ad copy.
Amazon is expanding its AI-powered Shop Direct program by letting merchants connect their product catalogs through feeds from platforms like Feedonomics, Salsify, and CEDCommerce. This allows brands that don’t sell directly on Amazon to appear in Amazon search results and in Rufus, Amazon’s AI shopping assistant. Shoppers can either click “Shop Direct” to buy on the brand’s website or use “Buy for Me,” where Amazon’s AI completes the purchase for them. The feed keeps catalog data, pricing, and inventory updated automatically, making it easier for brands to show products on Amazon without setting up a full marketplace store.
Why it matters:
Amazon is slowly turning its search engine into a broader shopping discovery layer. For sellers already on Amazon, this means more competition may appear in search results from brands that never list inventory on Amazon at all. For brands outside Amazon, it creates a lower-risk way to capture Amazon demand while still keeping the transaction and customer relationship on their own site.
Seller Central Update of the Week
Amazon added a small but useful upgrade to Product Opportunity Explorer this week. Sellers can now save product niches and ASIN opportunities directly inside the tool, creating a running list of ideas instead of restarting research every time. Saved opportunities appear in a new “Saved & Recent Views” tab and can be tracked across Amazon’s global marketplaces.
The signal is subtle but important. Amazon is steadily turning Seller Central into a full research environment, pulling workflows that once lived in tools like Helium 10 and Jungle Scout back inside its own ecosystem. Over time, product discovery and demand validation may increasingly happen directly on Amazon rather than through external software.
Tweet Spotlight
Meme Therapy
How did you like this week's newsletter?
Best,