Walmart's experiment with agentic checkout inside ChatGPT has produced data that few in the industry will find comfortable. Conversion rates for products sold directly inside the chatbot were three times lower than those requiring users to click out to Walmart's website, according to Daniel Danker, the executive vice president who oversees design and product at the retail giant. That figure, disclosed to WIRED in a report published March 18, 2026, represents one of the first hard performance disclosures from a major retailer testing AI-mediated commerce since the concept took hold in 2025.

The announcement matters. Walmart was the original launch partner for OpenAI's Instant Checkout feature, not a secondary participant. Its willingness to go on record with a critique of the product that it helped launch signals something meaningful about the actual state of consumer behavior in chatbot-driven commerce.

What Instant Checkout was, and what went wrong

OpenAI launched Instant Checkout on September 29, 2025, partnering with Stripe and initially enabling transactions from Etsy and Shopify merchants through ChatGPT. The infrastructure relied on the Agentic Commerce Protocol, co-developed with Stripe as an open standard for AI-mediated transactions. Walmart joined as a retail partner, making approximately 200,000 products available directly inside chat responses. Shoppers could provide payment and shipping details to OpenAI and complete purchases without leaving the ChatGPT interface.

That architecture, however logical in theory, clashed with how consumers actually shop. The single-item checkout model created a structural mismatch with typical retail behavior. Danker put it plainly: "They fear that when checkout happens automatically after every single item that they're going to receive five boxes when they actually just want it all in one." Shoppers did not want to split a checkout experience that separated one chatbot-purchased item from the other products already sitting in their Walmart cart.

The product categories that fared best within the Instant Checkout experience reveal a specific consumer profile. Top sellers included vitamin and protein supplements, particularly among users who had asked ChatGPT about GLP-1 weight-loss drugs and received guidance to increase nutrient intake. Other successful items tended to carry high enough price points to avoid triggering added shipping or small-basket fees. Collectively, the automotive, beauty, home management, hardware, and tools categories accounted for over half of all Instant Checkout orders. For higher-consideration products - a television, for instance - Walmart already knew the single-item flow was problematic; buying a TV without being prompted to add HDMI cables produces a frustrating installation experience.

Skepticism about the commercial viability of AI shopping agents had been building since October 2025, when independent analyst Andrew Lipsman published a detailed examination of eight structural challenges facing autonomous shopping systems. Among those challenges: consumer preference for seeing options before deciding, the complexity of return rate logistics, and the fundamental misalignment between retailer incentives and third-party AI intermediation. The Walmart data now provides a concrete data point that aligns with those concerns.

The Sparky solution: a chatbot inside a chatbot

Rather than spending years attempting to fix the "unsatisfying" consumer experience, as Danker described the Instant Checkout setup, OpenAI chose to move quickly toward a different architecture. Starting the week of March 25, 2026, Walmart's own chatbot - Sparky - will begin operating within ChatGPT. According to Danker, a similar integration will arrive in Google's Gemini chatbot the following month.

The structural change is significant. Instead of a checkout process embedded in ChatGPT's interface using OpenAI's payment infrastructure, consumers will interact with Sparky - effectively a Walmart-controlled environment running inside another company's chat interface. Baskets from Walmart's website, the Walmart app, and ChatGPT will synchronize with each other, allowing purchases to reflect ongoing shopping behavior rather than a single isolated transaction.

Sparky was developed by Walmart, according to Danker, but it runs on a combination of open source generative AI models alongside retail-specific models trained on decades of Walmart data. "We're able to route certain questions to one model and certain questions to another because we find that the quality of answers differs," Danker said. The architecture is intentionally flexible - designed to operate within external environments and adapt visually to feel like a native part of them. "It can take on slight tweaks to the look and feel, to make it feel like a natural part of other environments," Danker said.

The basket synchronization feature addresses what Danker identifies as the core problem with Instant Checkout. People add peanut butter on the app one day, aluminium foil the next, and a birthday gift on the website the evening before checking out. Forcing a separate single-item purchase inside ChatGPT broke that pattern. "When Sparky travels, it's the Walmart store meeting you where you are, instead of a completely broken experience," Danker said.

ChatGPT is now outpacing search engines for new customer acquisition

One data point from Danker deserves particular attention from the marketing community. ChatGPT is now bringing Walmart approximately twice the rate of new customers compared to search engines. Danker attributes this partly to the demographic profile of heavy ChatGPT users - they tend not to be typical Walmart customers. But Walmart's price positioning, product range, and geographic reach mean its inventory surfaces frequently in ChatGPT responses regardless of whether consumers are actively shopping.

For marketing professionals, this is a structural shift worth tracking carefully. OpenAI's shopping features have been expanding since April 2025, when the company introduced visual product browsing and reported over 1 billion web searches conducted through ChatGPT in a single week. The shopping functionality, which operates without sponsored placements, uses product metadata from third-party providers and intent-based signals to surface relevant items. That non-advertising model differs fundamentally from how Google Shopping or Amazon have historically operated.

OpenAI has since moved toward introducing advertising. In February 2026, the company began testing ads with a minimum brand commitment of $200,000, with early participants including Target, Ford, and Adobe, alongside holding company partners WPP Media, Omnicom, and Dentsu. The advertising layer sits on top of a commerce infrastructure that is, as the Instant Checkout data demonstrates, still working through its early-adoption difficulties.

Where Walmart's broader AI commerce strategy fits

Walmart has been moving aggressively across multiple AI commerce fronts. On January 11, 2026, Target and Walmart enabled direct purchasing through Google's Gemini AI assistant and AI Mode in Search, using the Universal Commerce Protocol - an open standard co-developed by the retailers and Google. That integration allows consumers to browse Walmart inventory and complete purchases without leaving Gemini, with account linking enabling personalized recommendations based on past online and in-store purchases.

Walmart Connect, the retail media arm, announced AI agent-powered advertising tools on January 6, 2026, including a conversational assistant for brands managing Sponsored Search campaigns. Walmart's global advertising business reached $6.4 billion in fiscal year 2026, a 46% increase, with Walmart Connect in the United States up 41% in the fourth quarter. Sparky users spend approximately 35% more per order than other shoppers, according to Walmart US CEO David Guggina - a figure that makes the chatbot's commerce performance central to the company's advertising economics, since larger baskets create more potential inventory for Sponsored Products placements.

The Sparky integration into ChatGPT and Gemini is partly a consequence of those economics. Walmart cannot afford for the AI-mediated shopping experience to underperform. Danker acknowledges that Sparky is currently slow and produces weak responses often enough that some consumers dismiss it as unreliable. The priority for 2026 is making Sparky more proactive, training it to know more about individual shoppers, and extending its competence across more departments, including pharmacy.

Half of Walmart app users have already engaged with Sparky, according to the company. Typical app users search for staples such as milk and bananas, but Sparky conversations tend toward more complex queries - exotic ingredients, solution-seeking for home projects, requests that span multiple departments.

A pivot that reflects broader industry dynamics

The shift from Instant Checkout to a Sparky-within-ChatGPT model fits a broader pattern in how the AI commerce ecosystem has evolved since September 2025. Microsoft launched Copilot Checkout on January 8, 2026, enabling direct purchases within the Copilot interface with support from PayPal, Shopify, and Stripe. Microsoft's internal data claimed that shopping journeys involving Copilot led to 53% more purchases within 30 minutes and 194% higher conversion when clear shopping intent was present - figures the Walmart data would suggest require careful scrutiny.

Major merchants, including most large UK and US retailers, have broadly welcomed AI agents, viewing them as a new customer segment to understand and optimize for, rather than a threat to block. Amazon is the primary exception, having filed a federal lawsuit against Perplexity AI in November 2025 for allegedly deploying covert agents into its platform without authorization. Danker is explicit that Walmart has taken the opposite stance: "We don't want to block things on a speculative or hypothetical concern." Walmart will support whatever tools customers use, as long as orders are accurate and the experience avoids erroneous purchases, shocking bills, or excessive customer service burden.

The Universal Commerce Protocol, which generated debate about surveillance pricing when it launched, creates a standardized technical layer connecting AI agents to retailer inventory, pricing, and checkout - using REST and JSON-RPC transport with support for Agent Payments Protocol, Agent2Agent, and Model Context Protocol. Its adoption by Walmart, Target, Shopify, and Etsy indicates that major commerce players are betting on open standards rather than proprietary integrations for the next phase of AI-mediated purchasing.

Mastercard and Google introduced Verifiable Intent on March 5, 2026, a cryptographic trust standard designed to create proof of consumer authorization at the moment an AI agent initiates a transaction. That infrastructure development, co-developed with IBM, Worldpay, Fiserv, Adyen, and others, addresses one of the structural concerns underpinning the Instant Checkout failure: consumer reluctance to hand payment authority to third-party AI systems they do not fully trust.

What the conversion data means for performance marketers

Three-times-lower conversion inside a chatbot compared to a standard click-out link is a substantial gap. It suggests that the friction of a familiar checkout flow - clicking through to a known retail website, seeing the full cart, and completing a purchase in an interface consumers have used before - provides meaningful reassurance that an embedded AI-mediated flow does not replicate.

OpenAI's announcement in September 2025 positioned the Agentic Commerce Protocol as enabling trust-focused design, with explicit user confirmation required before any action, encrypted payment tokens authorized only for specific amounts and merchants, and minimal data sharing. In practice, the data Walmart disclosed suggests those safeguards did not overcome consumer hesitation about buying through a chatbot.

For marketers planning investments in agentic commerce channels, the Walmart figure establishes a benchmark. Meta phased out Facebook and Instagram native checkout by August 2025, transitioning merchants back to website-based checkout on the grounds that it gave businesses greater control. The parallels with OpenAI's Instant Checkout pivot are direct: in both cases, platform-embedded checkout underperformed expectations, and the solution moved purchasing back toward merchant-controlled environments - albeit now in a new form through the Sparky architecture.

Whether the Sparky-inside-ChatGPT model performs better than the original Instant Checkout will take months to assess. Danker himself offered a measured view of how far AI commerce automation will actually reach: "This idea that it will all become automated might be a little bit far-fetched. People do get excited about shopping for clothes, for their home, for their children."

The basket synchronization approach, the chatbot-within-chatbot architecture, and the planned Gemini integration represent Walmart's best current answer to that challenge. What they do not represent is evidence that agentic commerce has solved the fundamental problem the Instant Checkout data exposed.

Timeline

Summary

Who: Walmart, represented by executive vice president Daniel Danker, who oversees design and product; OpenAI, represented by spokesperson Taya Christianson; and the broader ecosystem of AI commerce platforms including Google, Microsoft, and Amazon.

What: Walmart has disclosed that OpenAI's Instant Checkout feature - which allowed purchases to be completed directly inside ChatGPT - produced conversion rates three times lower than experiences requiring users to click through to Walmart's website. In response, OpenAI is discontinuing the Instant Checkout approach in favor of embedding Walmart's own Sparky chatbot inside ChatGPT, with basket synchronization across Walmart's app, website, and the ChatGPT interface. A parallel Gemini integration is planned for April 2026.

When: The Instant Checkout feature launched in November 2024 and ran for approximately five months. The performance data and the announcement of the Sparky pivot were published by WIRED on March 18, 2026. The Sparky-within-ChatGPT integration is scheduled to begin the week of March 25, 2026.

Where: The integration operates within ChatGPT's consumer interface, accessible to Plus, Pro, and Free users in the United States. The Gemini integration will extend the same approach to Google's AI assistant. Walmart's physical and e-commerce operations, spanning over 4,600 US stores, provide the inventory and fulfillment infrastructure.

Why: The single-item checkout model inside ChatGPT conflicted with typical multi-item shopping behavior. Consumers were reluctant to split purchases across multiple sessions and did not trust the embedded payment flow at the same rate as Walmart's own checkout. Walmart's motivation to fix this is commercial: ChatGPT now brings in new customers at twice the rate of search engines, making the quality of the AI-mediated shopping experience central to customer acquisition strategy.

Share this article
The link has been copied!