Walmart Abandons ChatGPT Native Checkout Over Poor Conversions

Walmart found that customers were three times less likely to complete a purchase inside ChatGPT compared to its own website. As a result, the retailer is abandoning OpenAI's 'Instant Checkout' in favor of its own integrated chatbot, Sparky. This shift suggests that keeping shoppers within a brand's owned digital environment remains the most effective strategy for driving conversions.

Key Points

• Walmart's test of 200,000 items in ChatGPT showed conversion rates were only one-third of those seen on its main website.
• OpenAI is officially phasing out its 'Instant Checkout' feature, which allowed users to buy products without leaving the AI interface.
• Walmart executives found the native AI shopping experience 'unsatisfying' for customers compared to traditional web shopping.
• The retailer is pivoting to an integrated model using its own 'Sparky' chatbot to bridge the gap between AI platforms and its owned web environment.
• A similar integration with Google Gemini is planned for the near future to maintain control over the checkout process.

Sentiment

The overall sentiment is skeptical and critical of AI-powered checkout. The community broadly agrees with the article's premise that ChatGPT checkout underperforms traditional e-commerce, viewing the result as unsurprising given the complexity of commerce infrastructure and the trust deficit inherent in AI-mediated purchasing. While a minority sees long-term potential, the dominant view is that this represents a fundamental mismatch between chat interfaces and optimized shopping experiences.

In Agreement

• E-commerce has been optimized over decades and OpenAI cannot easily replicate that institutional knowledge around checkout flows, trust, and catalog management
• ChatGPT checkout suffers from practical failures like showing out-of-stock items as available and linking to incorrect or outdated product listings
• OpenAI is primarily focused on model development and not genuinely interested in mastering the complexities of commerce infrastructure
• Chat-based shopping adds friction and trust concerns rather than removing them — it is a solution looking for a problem
• Consumer trust is fundamentally broken when an AI intermediary handles purchasing, especially given concerns about hallucinated product recommendations and unreliable inventory data

Opposed

• Early failures do not predict long-term outcomes — Netflix, Amazon, and PayPal all faced similar skepticism before succeeding, and AI shopping may follow the same trajectory
• Younger generations already use ChatGPT as their primary internet interface, suggesting the market will shift generationally regardless of current conversion rates
• LLMs as general-purpose tools could theoretically subsume entire product categories by erasing the boundaries that define software products, including e-commerce
• Natural language shopping could be genuinely better than clicking through bloated commerce sites, especially for niche or technical product searches
• A user-centric AI shopping agent acting in the buyer's interest could outperform existing systems by comparison shopping across retailers and optimizing for price