The Reality Check for Agentic Commerce
For the past year, the tech world has been buzzing with the promise of “agentic commerce”—a future where artificial intelligence doesn’t just suggest products but actually handles the entire transaction for you. The vision was simple: you tell ChatGPT you need ingredients for a dinner party or a new set of power tools, and the AI handles the search, the selection, and the checkout without you ever leaving the chat interface.
However, recent data from Walmart, the world’s largest retailer, suggests that we are much further from that reality than many anticipated. In a revealing disclosure, Walmart confirmed that conversion rates for purchases made directly inside ChatGPT were three times lower than when users were directed to Walmart’s own website.
This massive gap in performance highlights a critical friction point in the evolution of AI-driven shopping. While AI is excellent at discovery and curation, it is currently struggling to close the deal. For marketers, SEO professionals, and e-commerce platform owners, Walmart’s experience serves as a vital case study in why the “owned environment” still reigns supreme in the digital economy.
Inside the Experiment: Walmart and OpenAI’s Instant Checkout
The experiment began in earnest in November, when Walmart partnered with OpenAI to pilot a feature known as “Instant Checkout.” The initiative offered roughly 200,000 products that could be purchased natively within the ChatGPT interface. The goal was to remove the friction of jumping between apps and websites, creating a seamless “conversational” shopping experience.
On paper, it seemed like a win-win. OpenAI could demonstrate the utility of its ecosystem for commerce, and Walmart could reach tech-forward consumers exactly where they were spending their time. However, the results were far from the revolutionary breakthrough both companies hoped for.
Daniel Danker, Walmart’s Executive Vice President of Product and Design, did not mince words when describing the outcome. He noted that the in-chat purchases converted at only one-third the rate of traditional click-out transactions. More tellingly, Danker described the native AI checkout experience as “unsatisfying.”
Why Instant Checkout Failed to Convert
To understand why a 3x difference in conversion exists, we have to look at the psychology of the modern shopper and the technical limitations of current LLM (Large Language Model) interfaces.
1. The Lack of Visual Richness
E-commerce is a visual medium. When a user visits Walmart.com, they are greeted with high-resolution images, video demonstrations, 360-degree product views, and detailed size charts. ChatGPT, by its nature, is a text-heavy interface. While it can display images, the rich, interactive experience of a dedicated retail site is difficult to replicate in a scrolling chat window. Shoppers often need that final visual confirmation before hitting “buy,” a step that feels less certain inside a third-party AI tool.
2. The Trust and Security Gap
Entering credit card information and personal shipping details into a chatbot feels fundamentally different from doing so on a brand’s official website. Despite the security protocols in place, there is a lingering “trust gap” when it comes to agentic commerce. Consumers are comfortable asking ChatGPT for a recipe or a summary of a news article, but trusting it to handle a financial transaction with a third-party retailer introduces a new layer of hesitation.
3. Missing Social Proof and Nuance
Walmart’s website is optimized for conversion through social proof—reviews, ratings, and “customers also bought” suggestions. While an AI can summarize reviews, the raw data of seeing thousands of verified purchases and reading specific user feedback provides a level of reassurance that a summarized AI response lacks. If the AI says, “This is a highly-rated drill,” it carries less weight than seeing 5,000 four-star reviews on the product page itself.
The Death of OpenAI’s Instant Checkout
The disappointing results from the Walmart partnership have had immediate consequences for OpenAI’s product roadmap. Earlier this month, OpenAI confirmed that it is phasing out the “Instant Checkout” feature entirely.
This pivot marks a significant shift in how AI labs view commerce. Rather than trying to be the “everything store” that manages transactions internally, OpenAI is moving toward a model where the AI acts as a sophisticated lead generator, handing the final transaction back to the merchant.
This is a victory for the traditional web and for brand-owned platforms. It suggests that for the foreseeable future, the “buy” button belongs on the retailer’s site, not in the LLM’s sidebar.
Enter Sparky: Walmart’s New Strategy for AI Integration
Walmart isn’t abandoning AI; it is simply changing how it integrates with it. The company is moving away from native ChatGPT checkouts and toward an “embedded” model. This involves the deployment of “Sparky,” Walmart’s own proprietary AI shopping assistant.
Instead of a generic OpenAI checkout process, Sparky will be embedded within the ChatGPT ecosystem. This new approach changes the dynamic in several key ways:
Syncing the Shopping Experience
One of the biggest frustrations with the previous model was the lack of continuity. In the new version, users will log into their Walmart accounts through the interface. This allows for cart syncing across platforms. If you add an item to your cart via a conversation in ChatGPT, it will appear in your Walmart app and on the Walmart website. This creates a “persistent cart” that bridges the gap between AI discovery and traditional checkout.
Merchant-Handled Transactions
By moving the checkout back into Walmart’s system—even if it is triggered from within ChatGPT—Walmart regains control over the user experience. They can ensure that shipping options, loyalty points (like Walmart+), and promotional offers are applied correctly. This “app-based checkout” model is what OpenAI is now favoring for all its merchant partners.
Multi-Platform Presence
Walmart isn’t putting all its eggs in the OpenAI basket. The company confirmed that a similar integration with Google Gemini is slated for next month. By treating AI platforms as distribution channels rather than transaction hubs, Walmart is positioning itself to be present wherever the consumer starts their search journey.
What This Means for SEO and Digital Marketing
The Walmart/OpenAI data is a wake-up call for the SEO and digital marketing industry. For years, there has been a growing fear that “Search Generative Experience” (SGE) and AI bots would “steal” traffic from websites, keeping users in a walled garden where they never see the original source.
Walmart’s 3x conversion disparity suggests that even if AI bots can provide answers, they cannot yet provide the “experience” required to drive revenue at scale. Here are the key takeaways for digital strategists:
1. Owned Media is More Important Than Ever
If a user is 300% more likely to buy on your website than in a chat interface, your website remains your most valuable asset. The focus should not just be on “ranking” in AI responses, but on ensuring that the transition from an AI response to your landing page is as frictionless as possible.
2. Optimization for “Handoffs”
The new goal for SEO might not just be “clicks,” but “successful handoffs.” As Walmart integrates Sparky into Gemini and ChatGPT, the metric for success becomes how effectively the AI can move a user into the merchant’s logged-in ecosystem. Technical SEO will need to evolve to ensure that deep-linking and cart-syncing APIs are working perfectly.
3. The Role of Brand Authority
The reason users click out to Walmart.com is that they know and trust the Walmart brand. In an AI-driven world, brand signals become even more critical. If an AI mentions a generic product versus a branded product, the branded product will almost always have a higher conversion rate because of the pre-existing trust. Building a brand that users *want* to visit directly is the best defense against AI-driven disintermediation.
The Future of Agentic Shopping
While Walmart’s current data shows a preference for traditional websites, we shouldn’t assume agentic commerce is dead. We are currently in the “Beta phase” of AI shopping. As LLMs become more multimodal—incorporating voice, video, and real-time data—the gap may eventually narrow.
However, the “unsatisfying” nature of current AI shopping highlights a fundamental truth about retail: it is not just a utility; it is an experience. Shopping involves browsing, comparing, and the psychological satisfaction of interacting with a brand.
Walmart’s shift to using Sparky as a bridge rather than a standalone store is a pragmatic move. It acknowledges that AI is a powerful tool for discovery, but that the retailer’s own environment is where the value is truly realized.
Conclusion: The Balanced Ecosystem
The revelation that ChatGPT checkout converted 3x worse than Walmart’s website is a defining moment in the current AI cycle. it proves that for all the intelligence of LLMs, they have not yet mastered the art of the sale.
For Walmart, the path forward is clear: utilize AI to enhance the search and discovery phase, but bring the customer back home to complete the journey. For the rest of the tech and gaming world, the lesson is equally clear: don’t be too quick to abandon your owned platforms in favor of the latest AI trend. The website is not dead; in fact, its role as the ultimate conversion engine has never been more secure.
As we look toward the integration of Sparky in Google Gemini and the continued evolution of ChatGPT, the industry will be watching closely. If Walmart can successfully bridge the gap between AI conversation and website conversion, they will have created a blueprint for the future of retail in the age of artificial intelligence. For now, however, the “click-out” remains the king of the transaction.