Walmart Pivots Away from OpenAI's Instant Checkout After Poor Conversion Rates, Shifts to Embedded Sparky Chatbot
Summary
- • Walmart's Instant Checkout inside ChatGPT had conversion rates three times lower than click-out links
- • Walmart's Sparky chatbot will embed directly inside ChatGPT starting next week
- • ChatGPT now brings Walmart twice as many new customers as search engines
- • Sparky will sync shopping carts across Walmart app, website, and ChatGPT sessions
Details
Instant Checkout conversion rates were 3x lower inside ChatGPT than click-out links
Walmart EVP Daniel Danker disclosed this directly to Wired. The data point is the clearest public evidence yet that embedding checkout inside AI chat does not improve purchase rates and may actively hurt them compared to redirecting users to a retailer's own site.
Walmart's Sparky chatbot will operate inside ChatGPT starting next week
Rather than OpenAI handling checkout directly, Walmart's own AI assistant Sparky will be embedded as a chatbot-within-a-chatbot. A similar integration with Google Gemini is planned for the following month. This gives Walmart direct control over the shopping experience within third-party AI platforms.
OpenAI's Instant Checkout agentic commerce feature is being deprecated in favor of the embedded chatbot model
OpenAI launched Instant Checkout with Walmart, Etsy, and other retailers as a commission-based revenue model. The pivot away from it — driven by poor results — represents an early and significant setback for the agentic commerce thesis that AI platforms can own the transaction layer.
Single-item auto-checkout conflicts with consumers' multi-session basket-building habits
Danker identified the core UX problem: shoppers fear receiving multiple separate shipments and dislike having a purchase execute automatically before they have finished assembling their full cart. Cart sync across Walmart's app, website, and ChatGPT is the proposed fix.
ChatGPT now brings Walmart roughly twice the new-customer rate that search engines do
Danker attributes this partly to ChatGPT's power-user demographic skewing away from typical Walmart shoppers. Walmart's scale — price, selection, and geographic reach — means its products surface frequently in ChatGPT responses regardless, making the channel valuable for acquisition even if conversion is low.
Walmart made about 200,000 products available via Instant Checkout inside ChatGPT
Higher-ticket items like TVs were excluded and required clicking out to Walmart's website. Top-selling categories in Instant Checkout included vitamins, protein supplements, automotive, beauty, home management, hardware, and tools — which together accounted for over half of all orders.
Sparky combines open-source generative AI with retail-specific models trained on decades of Walmart data
Walmart built Sparky in-house. The system dynamically routes questions to different models depending on which produces the best answer quality for that query type. This multi-model routing approach gives Walmart flexibility without being locked into a single AI provider.
Walmart's goal is to make Sparky a portable store presence across AI platforms
Danker framed the embedded chatbot model as 'the Walmart store meeting you where you are.' The simultaneous rollout to ChatGPT and Gemini suggests Walmart intends to distribute Sparky across multiple AI interfaces rather than betting on one platform.
GLP-1 drug users asking ChatGPT about weight-loss medications are driving supplement sales
ChatGPT responses advising GLP-1 users to increase nutrient intake have driven supplement sales through Instant Checkout. This illustrates how AI health conversations can generate downstream retail demand — a dynamic Walmart is positioned to capture given its supplement catalog.
Market Impact = commercial performance results, Product Launch = new product or feature going live, Industry Update = structural change to a business model or program, Insight = analytical observation about behavior or trends, Stat = specific quantitative data point, Tech Info = technical architecture or capability detail, Strategy = deliberate business positioning or directional decision, Context = background information that frames the story
What This Means
The failure of OpenAI's Instant Checkout is an early reality check on agentic commerce: giving an AI platform control over the transaction layer does not automatically improve conversion, and may make shopping worse by clashing with how people actually build and complete purchases over time. Walmart's pivot to embedding its own Sparky chatbot inside ChatGPT and Gemini suggests the smarter path is retailers bringing their own AI into third-party platforms — keeping cart continuity, brand context, and data in-house — rather than outsourcing the purchase experience to the platform. For the broader industry, this signals that AI chat may prove more valuable as a discovery and acquisition channel than as a direct checkout mechanism, at least in its current form.
Sentiment
Broadly reporting Instant Checkout flop as OpenAI setback, supportive of Walmart's pivot to Sparky
“$WMT is disappointed in results from OpenAI partnership... Instant Checkout has been a flop.” -- Wired. OpenAI on a heater recently in the news…”
“$WMT is reportedly disappointed with its OpenAI shopping partnership as ChatGPT’s Instant Checkout has underperformed expectations. In-chat purchases are converting at roughly one-third the rate of purchases made directly on Walmart’s site.”
“Walmart just learned that owning the decision layer matters more than owning checkout inside someone else’s interface.”
Framing pivot as key lesson for retailers
“the problem: Instant Checkout bought items one at a time. but users already had full carts sitting in the #Walmart app... fix: Sparky embeds into #ChatGPT next week... full cart management in one place instead of item-by-item.”
Split
Strong consensus on flop (~90%), split on OpenAI impact: major agentic commerce setback (~70%) vs. pragmatic evolution (~30%).