# Walmart Ditches Instant Checkout in ChatGPT Due to Low Conversion Rates
Walmart has released conversion data for purchases via ChatGPT: it's three times lower than redirecting to their own site. The company is shutting down the Instant Checkout feature, launched in November 2025, and shifting to integrating their own AI assistant Sparky. OpenAI is also ending support for Instant Checkout, handing order management back to sellers.
Instant Checkout allowed checkout for around 200,000 items directly in the chatbot interface without going to the Walmart site. VP of Product Daniel Denker described this approach as unsatisfactory. Key issues included integration with loyalty programs, tax calculations, returns processing, and anti-fraud measures.
Shifting to an Embedded Assistant Model
The new strategy involves embedding Sparky into ChatGPT. Users will be able to log into their Walmart account and sync their cart across the site, app, and chatbot. This brings the Walmart store into the OpenAI ecosystem, rather than the other way around.
A similar implementation is planned for Google Gemini in April 2026. Sparky has already proven effective:
- Users spend 35% more when interacting with the assistant.
- About 50% of Walmart mobile app users use it.
- ChatGPT generates twice as many new customers compared to search engines.
This audience differs from the retailer's traditional customer base, making the integration promising.
Challenges of Agentic Commerce
Instant Checkout became the first public case of failure for direct commerce in chatbots from a major retailer. The agentic model, where AI fully handles the transaction, doesn't account for e-commerce specifics:
- Syncing loyalty and personalization.
- Dynamic tax and shipping calculations.
- Handling bundles and promotions.
- Integration with anti-fraud systems.
- Returns and refunds policies.
OpenAI couldn't abstract these processes, leading to abandoning the model. The question is whether the embedded approach will solve these challenges without sacrificing UX.
Outlook for AI Integration Developers
For teams working with OpenAI and Google Gemini APIs, the Walmart case highlights risks of full-agent commerce. Focus on hybrid models is recommended:
- Authorization and Sync: OAuth 2.0 for seamless login, WebSocket or polling for cart.
- Fallback Redirect: For complex transactions, redirect to native checkout.
- Success Metrics: Conversion, AOV (average order value), CAC (customer acquisition cost).
Walmart's data shows branded assistants boost retention by 35%. Integration with external platforms requires strict data ownership control.
Key Takeaways
- Instant Checkout conversion is three times lower than Walmart's site.
- Sparky increases average order value by 35%, used by 50% of app users.
- ChatGPT brings in twice as many new customers as search engines.
- OpenAI is winding down direct checkouts, focusing on seller-managed.
- The failure highlighted challenges with loyalty, taxes, and anti-fraud in agent commerce.
— Editorial Team
No comments yet.