If you sell anything through Walmart, or you compete with brands that do, the way people find your product is quietly changing. Someone can now type "best mattress under $1,000" or "gift for an avid reader" into ChatGPT and get a recommendation, a comparison, and a path to checkout, without ever landing on walmart.com or seeing a search results page. That's not a hypothetical. It's live, and it's already been through one major rewrite in under six months.
Here's the actual story of the Walmart and OpenAI partnership, what changed between the launch and now, and what it means if your brand's revenue runs through Walmart's shelves, physical or digital.
What Walmart and OpenAI actually announced
On October 14, 2025, Walmart announced a partnership with OpenAI to bring what both companies called "AI-first shopping experiences" to ChatGPT. The headline feature was Instant Checkout: customers could describe what they wanted in plain language, get product options pulled from Walmart's catalog, and complete the purchase without leaving the chat window.
Sam's Club members got the same access under the same integration. Walmart CEO Doug McMillon framed the shift in blunt terms: "eCommerce shopping experiences have consisted of a search bar and a long list of item responses. That is about to change." The pitch was "agentic commerce," a term you'll see everywhere in retail decks for the next two years. The idea is that shopping stops being reactive (you search, you scroll, you decide) and becomes proactive: the system learns what you buy, predicts what you'll need, and surfaces it before you go looking.
Walmart also used the announcement to talk about internal AI adoption, rolling out ChatGPT Enterprise to associates and citing existing wins like cutting fashion production timelines by 18 weeks and trimming customer care resolution times by 40%, according to its own reporting.
Instant Checkout itself wasn't new when Walmart joined. OpenAI had launched it about a month earlier, on September 29, 2025, with Etsy as the first live merchant, built on the Agentic Commerce Protocol, an open standard OpenAI co-developed with Stripe. PayPal joined as a second payment rail shortly after. Walmart was one of the first big-box retailers to bolt onto that rail rather than build its own competing standard, which mattered more than it looked at the time.
The pivot nobody expected: Sparky replaces in-chat checkout
Five months later, the deal looked different. In March 2026, Walmart and OpenAI reworked the integration, and the change was not cosmetic. Instead of completing a Walmart purchase entirely inside ChatGPT's own checkout flow, Walmart plugged in Sparky, its own generative AI shopping assistant, as an embedded app experience within ChatGPT.
Here's the practical difference. Product discovery still happens in the chat. But once you're ready to buy, ChatGPT now hands you off to a Walmart-branded environment for account linking, loyalty, and payment, rather than closing the loop through OpenAI's original checkout. OpenAI was candid about why: "the initial version of Instant Checkout did not offer the level of flexibility that we aspire to provide, so we're allowing merchants to use their own checkout experiences while we focus our efforts on product discovery." Walmart's Daniel Danker put the retailer's read on it this way: "Shopping can start anywhere now, whether that's Walmart or a question in ChatGPT."
Translation for anyone who isn't fluent in corporate-speak: OpenAI decided it doesn't want to own the messy, retailer-specific parts of checkout (loyalty programs, stored payment methods, return policies, tax logic). It wants to own discovery, the moment where the AI decides which product to mention. Walmart, and every other retailer that followed, wants to keep control of the transaction and the customer relationship once that moment happens.
That's a meaningfully different power split than the original October announcement implied, and it's the part most coverage glossed over. Walmart didn't leave ChatGPT. It renegotiated for more control over its own customers.
Who else moved on the same protocol
The Agentic Commerce Protocol didn't stay a Walmart-and-Etsy story. By early 2026, Target, Sephora, Nordstrom, Lowe's, Best Buy, The Home Depot, and Wayfair had all integrated with the updated discovery-first version of ACP. Walmart separately struck a parallel deal with Google to plug its catalog into Gemini's shopping surfaces, which tells you this isn't an OpenAI-exclusive bet for Walmart either. It's hedging across every major AI assistant that might become a shopping front door.
Here's how the two models compare in practice:
| Instant Checkout v1 (Sept-Oct 2025) | Discovery-first ACP (March 2026 onward) | |
|---|---|---|
| Where discovery happens | Inside ChatGPT | Inside ChatGPT |
| Where checkout happens | Inside ChatGPT, via OpenAI's flow | Handed off to the retailer's own environment (Sparky, in Walmart's case) |
| Who owns loyalty and payment data | Split between OpenAI's checkout layer and the merchant | The retailer, fully |
| First merchant | Etsy | Walmart, Target, Sephora, Nordstrom, Lowe's, Best Buy, Home Depot, Wayfair |
| OpenAI's stated focus | End-to-end transaction | Product discovery and recommendation only |
The direction of travel is clear. OpenAI is positioning ChatGPT as the place where the shopping decision gets made, not necessarily the place where the money changes hands. That's a narrower, more defensible role, and it's probably the more durable one.
What this signals about retail's shift to agentic commerce
Three things stand out if you zoom out from the Walmart specifics.
- The "front door" of ecommerce is fragmenting. For twenty years, that front door was a search bar, either Google's or the retailer's own site search. Now there are multiple AI assistants (ChatGPT, Gemini, and increasingly Perplexity and Claude) that can intercept the shopping decision before a customer ever opens a retailer's app. Walmart hedging across both OpenAI and Google in the same quarter is the clearest signal yet that big retailers see this as inevitable, not experimental.
- Retailers are drawing a hard line at checkout. The pivot from Instant Checkout to Sparky's handoff model shows the largest merchants aren't willing to let a third party own the transaction layer, even one as dominant as OpenAI. Expect every future retailer-AI deal to follow this same discovery-in, checkout-out pattern rather than the original all-in-one vision.
- Being cited by the model now matters as much as ranking in search. If ChatGPT is the layer that decides which three products get mentioned when someone asks for "a gift for an avid reader," a brand's job isn't just SEO anymore. It's making sure the underlying product data (structured feeds, accurate pricing, current stock, clear specs) is something an AI model can parse and trust enough to recommend confidently.
What it means for brands sold through Walmart, and retailers generally
If your products live on Walmart's marketplace, or you're a retailer watching this play out and wondering whether to build toward it, the practical shift is this: your product listing is no longer just competing for a spot on a search results page. It's competing to be the answer an AI model gives when someone describes a need in plain language.
A few things follow from that, and none of them are optional if you want to show up in this channel:
- Clean, structured product data stops being a nice-to-have. Assistants pulling from a retailer's catalog through a protocol like ACP need accurate titles, specs, pricing, and availability in a machine-readable format. A listing with vague titles or missing attributes is invisible to an agent doing product matching, even if it ranks fine on walmart.com's own search.
- Reviews and third-party mentions carry more weight than they used to. When an assistant is deciding which product to recommend for "durable" or "best for small kitchens," it's drawing on more than the retailer's own metadata. Independent reviews, comparison content, and how your brand is discussed elsewhere on the web all feed into whether a model trusts it enough to surface.
- You need to know whether you're actually showing up. Most brands have no visibility into what ChatGPT, Gemini, or Perplexity say about them when a shopper asks a comparison question. That's a blind spot that didn't exist when the only channel was Google search rankings you could check yourself.
- The retailer relationship now includes an AI layer you don't control. Walmart, not individual brands, negotiates the ACP integration and decides what data gets exposed to OpenAI. Sellers on Walmart's marketplace should be asking their account managers directly whether their catalog data is flowing into these AI shopping surfaces and how it's represented there.
This is the same problem AI Peekaboo customers are already dealing with in search generally: you can't fix what you can't see. Tools built for tracking ChatGPT visibility exist for exactly this reason, and the same logic that applies to informational queries applies to shopping ones. If you want to understand how ChatGPT actually breaks a query like "best budget laptop for college" into sub-questions before it recommends anything, it helps to understand how fan-out queries work, since that's the mechanism deciding which of your product pages even gets considered.
For ecommerce brands more broadly, not just Walmart sellers, this partnership is a preview of what every major retailer will eventually have to negotiate with at least one AI platform. If you're running an online store today, the tools built specifically for ecommerce AI visibility are worth a look before your competitors get there first, because the brands that show up in these answers now are the ones building the structured data and citation profile for it today.
How Sparky and ChatGPT actually decide what to show you
Get precise about what's happening technically here, because "AI picks a product" undersells how mechanical most of this still is. When you ask ChatGPT something like "I need a durable backpack for a toddler," the model doesn't browse Walmart's site the way a person would. It queries structured product data exposed through the Agentic Commerce Protocol, which includes fields like title, price, availability, category, and merchant-supplied attributes. The model then matches those fields against your query and whatever it can infer from surrounding context, like budget signals or past turns in the conversation.
That means two products with identical quality but wildly different metadata completeness can get very different treatment. A listing with a generic title, no size or material attributes, and stale stock data is functionally invisible to this matching process, even if it's a five-star product on the shelf. Sparky works the same way inside Walmart's own environment, layering in Walmart-specific signals like purchase history and loyalty status once a shopper lands there after the ChatGPT handoff.
This is also why the discovery-first split matters more than it sounds. OpenAI keeping ownership of discovery means OpenAI's model, its training data, and its own sense of what counts as a good answer are the filter standing between your product and the customer. Walmart's Sparky only gets a shot at influencing the outcome after that filter has already run. If your product doesn't clear the ChatGPT-side matching step, it never reaches the point where Walmart's own recommendation logic could have surfaced it anyway.
For a brand manager, the uncomfortable takeaway is that you now have two AI systems to satisfy instead of one traditional search algorithm: whatever ChatGPT (or Gemini, if you're going through that channel) uses to decide what to mention, and whatever the retailer's own assistant does once a shopper arrives. Optimizing for just one leaves a gap in the other.
Where this is likely headed
I'd bet the discovery-in, checkout-out split becomes the industry default, not a Walmart-specific compromise. It solves the actual disagreement here: OpenAI wants the attention and the recommendation moment, retailers want the transaction and the customer data. Neither side is going to give that up, and this structure lets both keep what they actually want.
What's less settled is which AI platform ends up mattering most for shopping queries. Right now it's genuinely split between ChatGPT and Gemini, with Perplexity pushing into the same space. Betting your visibility strategy on just one of them would be a mistake. The brands that treat this the way they'd treat multi-platform SEO, tracking presence across all the assistants that might field a shopping question, are the ones that won't get caught flat-footed when the next protocol shift happens.
If you want a clearer picture of how your own products show up (or don't) across ChatGPT, Gemini, and Perplexity's shopping answers, email us at filipe@aipeekaboo.com or grab 30 minutes on the calendar. We'll walk through what an actual audit of your AI shopping visibility looks like.
