Walmart's AI shopping assistant, Sparky, is no longer just a feature inside the Walmart app. It now operates inside ChatGPT, inside Google Gemini, and in physical Walmart store locations. On Walmart's Q1 FY2027 earnings call, covering the quarter that ended April 30, 2026: CEO John Furner reported that Sparky's weekly active users grew more than 100% in a single quarter, that units purchased through Sparky increased more than fourfold from the prior quarter, and that customers who use Sparky have an average order value 35% higher than those who don't.
Those numbers got a lot of attention. What got less attention is the technical architecture behind how Sparky shows up in three very different environments, and why a brand seller's listing data determines whether Sparky can find them at all.
Sparky Native: The Walmart App and Walmart.com
The version of Sparky that Walmart has the most control over lives in the Walmart app and on Walmart.com, where it's accessible through an "Ask Sparky" interface. This is also the most capable version, because here Sparky has direct access to the full range of Walmart's internal data: your listing quality score, attribute completeness, conversion history, sales velocity, inventory levels, and real-time pricing. No external AI surface gets that depth of data.
When a shopper types something like "I need to plan a birthday party for a six-year-old," Sparky doesn't return a search results page. It interprets the intent, builds a product basket across relevant categories, and presents recommendations with explanations. It can handle reorders automatically for items a customer has purchased before, and as of this year it works in Spanish. Starting in June 2026, Walmart also deployed Sparky in physical stores, where customers can interact with it to find products, check stock, and place orders for pickup or delivery.
This is the surface where your listing optimization has the most direct impact. Sparky reads your attribute fields semantically: it's not matching keywords, it's matching meaning. A product with thin attributes and a generic description is invisible to Sparky even if it ranks fine on the keyword search page. The two systems use different signals.
The ChatGPT Integration: Why Walmart Walked Away from Instant Checkout
Walmart's ChatGPT story is worth understanding in detail, because the path they took says a lot about where agentic commerce is heading.
In 2025, Walmart partnered with OpenAI to offer what OpenAI called Instant Checkout: a shopper could ask ChatGPT about a product, and the transaction would complete right inside the chat interface, handled by OpenAI's payment layer. It sounded elegant. It didn't work well. Walmart disclosed in March 2026 that the conversion rate on Instant Checkout purchases was roughly one-third of what Walmart's own site achieved for the same products.
Customers found an AI surface that didn't know their loyalty program, their Walmart+ membership, their saved payment methods, or their purchase history. A checkout experience that doesn't know who you are converts poorly. This shouldn't surprise anyone.
Walmart's response was to replace that arrangement with something different. Starting the week of March 25, 2026, Walmart embedded its own Sparky agent directly inside the ChatGPT interface, taking over the experience at the point where a shopper moves from browsing to buying. When a user asks ChatGPT about a product and it surfaces something from Walmart's catalog, the handoff to Sparky brings the user into Walmart's own account-linking flow, loyalty program, and checkout system, while they're still nominally inside ChatGPT.
OpenAI becomes the discovery surface. Walmart controls the transaction. After the switch, conversion recovered to roughly 70% of on-site levels, according to reporting on the rollout.
This matters for sellers because it clarifies who is running the recommendation engine. When ChatGPT surfaces a Walmart product, it's pulling from data Walmart has made available through this integration, which means your Walmart listing data, not some outside crawl, determines whether you appear. If your attributes are incomplete, Sparky doesn't have what it needs to match you to conversational queries, and it won't.
The Google Gemini Layer: UCP and What It Really Means
The Gemini integration runs on different plumbing. In January 2026, at the National Retail Federation conference, Google announced the Universal Commerce Protocol (UCP), an open standard for agentic shopping co-developed with Shopify, Walmart, Target, Etsy, Wayfair, and more than 20 payment and retail partners including Mastercard, Visa, Stripe, Adyen, Best Buy, and The Home Depot.
UCP is worth understanding as a concept, because it's the clearest description of how agentic commerce works at a protocol level. Think of it as a shared language that lets an AI agent and a retailer's systems talk to each other across four layers: discovery (finding the right product), cart (building and holding a selection), checkout (completing the transaction), and post-purchase (order confirmation, tracking, returns). UCP defines how each of those steps gets communicated between an AI agent and a retailer's backend.
The underlying transport layer supports REST, MCP (Model Context Protocol), and Agent2Agent (A2A) protocols, which is why the concept feels familiar if you've been following how AI tools connect to external systems generally. It's not a marketplace. It's a specification for how agents and merchants connect.
For a shopper, the experience on Gemini looks like this: they ask Gemini to find a specific type of product, Gemini queries the UCP endpoints of participating retailers including Walmart, pulls live inventory and pricing, surfaces recommendations, and can complete checkout via Google Pay or PayPal, without the shopper ever visiting Walmart.com. Walmart's Sparky assistant is also deployable across Gemini surfaces separately from UCP, and the two layers work together. In July 2026, Walmart extended the Google Gemini integration to cover in-store point-of-sale checkout as well, meaning the same catalog data now routes through AI-mediated checkout across digital and physical environments simultaneously.
A critical distinction: external AI surfaces like Gemini only get what Walmart makes available through its feeds and UCP integration. They don't get Walmart's internal conversion data, loyalty data, or listing quality scores. That's another reason why the native Sparky experience in the Walmart app is the highest-signal surface, and why Walmart has a strong interest in steering shoppers toward its own interface rather than external ones.
What This Means for Your Listings
All three of these surfaces, native Sparky, Sparky-in-ChatGPT, and Gemini via UCP, draw from the same foundational source: your Walmart catalog data. Structured attributes, complete specifications, clear product descriptions written in natural language rather than keyword strings, Q&A content, and review density. These are the inputs that allow an AI agent to understand what your product is, who it's for, and when to recommend it.
Traditional search optimization on Walmart focused on keyword placement and category fit. Those still matter for the keyword search surface. But Sparky and the AI agents that pull from Walmart's catalog are running a different evaluation: can your listing answer a conversational question, not just match a keyword string.
A listing that says "12-piece stainless steel cookware set, induction compatible, oven safe to 500°F, dishwasher safe, includes lids" gives an AI agent something to work with. A listing that says "premium kitchen cookware nonstick professional grade" gives it almost nothing.
The 35% higher order value among Sparky users isn't happening because Sparky is randomly picking winners. It's happening because Sparky tends to surface products it can explain: products with enough structured data to generate a coherent recommendation. The brands that win on Sparky are, in large part, the brands whose listings make Sparky's job easy. You can read more about how we approach Walmart listing optimization as part of a broader channel strategy built for AI discovery.
If you want to work through what your Walmart catalog looks like through Sparky's eyes and close the gaps before Sparky's growth makes them more expensive to ignore, schedule a call and we'll start with your top-selling SKUs.