Walmart’s ChatGPT Checkout Converts 66% Worse Than Its Own Site, Raising AI Commerce Doubts

Why It Matters

The Walmart episode highlights a fundamental tension in e‑commerce: AI can dramatically expand product discovery, but trust and transactional control remain anchored to retailer‑owned experiences. A 66% conversion drop signals that shoppers still demand the familiar checkout environment, especially for high‑value or complex purchases. Retailers that can seamlessly blend AI‑powered discovery with their own secure checkout may capture new traffic without sacrificing conversion. For the broader retail sector, the shift from OpenAI’s Instant Checkout to Walmart‑controlled Sparky underscores a strategic pivot toward data ownership. By keeping the cart and checkout within its ecosystem, Walmart can gather early‑stage shopper intent, refine personalization, and potentially close the discovery‑conversion gap that Criteo’s survey shows exists for most consumers.

Key Takeaways

• Walmart’s Instant Checkout conversion was about one‑third of its website’s rate, a 66% shortfall.
• Roughly 200,000 Walmart products were available through OpenAI’s ChatGPT in the pilot.
• Walmart plans to embed its Sparky assistant inside ChatGPT and Gemini, shifting checkout back to its own platform.
• Criteo survey: 47% use AI for product comparison, but only 8% complete purchases inside AI tools.
• Andy Stephen (Criteo) says AI assistants must deliver "relevant surprises" to become the gold standard for shopping.

Pulse Analysis

Walmart’s retreat from OpenAI’s Instant Checkout is less a failure of AI technology than a reminder that commerce is as much about trust as it is about convenience. The 66% conversion gap mirrors a broader consumer reluctance to hand over payment details to a third‑party conversational interface, especially when the experience feels detached from the retailer’s brand. By re‑asserting control through Sparky, Walmart is betting that the perceived security of its own checkout will outweigh the friction of leaving the AI chat.

Historically, retailers that have attempted to outsource the checkout experience—think early mobile payment pilots—have struggled to achieve parity with native site performance. The current wave of "agentic commerce" repeats that pattern, but with a more sophisticated front‑end. The key differentiator will be data integration: if Sparky can surface personalized offers, real‑time inventory and seamless cart sync, it may close the gap that Criteo’s data shows exists between discovery and purchase. Conversely, any latency or data mismatch could reinforce shopper skepticism and cement the status quo of AI as a discovery layer only.

Looking ahead, the industry will watch Walmart’s rollout as a litmus test for the viability of hybrid AI‑retailer models. Success could trigger a cascade of similar integrations from other big‑box players, prompting AI platform providers to build more robust merchant APIs. Failure, however, may push the conversation back toward improving the hand‑off experience rather than attempting full in‑chat transactions, reaffirming the notion that the future of e‑commerce lies in a collaborative, not a monolithic, AI‑driven checkout ecosystem.