One number tells the whole story: Walmart's purchases through ChatGPT Instant Checkout are converting at roughly one-third the rate of purchases on Walmart's own site, according to a March 18 report from The Information. After months of hype around in-chat buying, that is not a soft miss. It is a channel problem with a hard number attached to it.
The disappointing part for Walmart is obvious. This was supposed to be one of the cleanest examples of AI-native commerce moving from demo to habit. Walmart had brand trust, massive catalog depth, fulfillment infrastructure, and a direct partnership with OpenAI. If conversational checkout was ready for mainstream retail, this was the kind of relationship that should have shown it.
Instead, the most interesting result is how ordinary the failure looks.
One-third is not a rounding error
Walmart and OpenAI framed the partnership around a simple idea: let people discover products, compare options, and complete the purchase without leaving ChatGPT. OpenAI's own Instant Checkout materials describe a flow where the user can tap Buy inside the chat experience and complete payment there instead of jumping out to the merchant's site. Walmart, when it announced the partnership, pitched the broader vision as AI-first shopping that would feel more contextual and convenient than the standard search-bar-plus-grid pattern.
On paper, that sounds inevitable. In practice, one-third conversion suggests the chat interface is failing at a job Walmart's own site already does better.
That distinction matters. A weak conversion rate is not just a UX nitpick. It means the new channel is introducing enough hesitation, confusion, or missing context that two out of every three conversions, relative to Walmart.com, are disappearing before payment clears. That is the kind of gap that forces executives to stop talking about the future and start asking whether the product is solving the right problem at all.
Friction hides inside the conversation
This outcome makes technical sense. Buying inside a chat thread sounds seamless only if you ignore what high-intent commerce sessions actually look like.
Walmart's site is built for commitment. The shopper can scan variants, compare shipping promises, inspect reviews, browse adjacent items, check return policies, see cart totals, and move through a purchase flow they already understand. The interface is boring in the useful way: it reduces ambiguity at the exact moment a customer wants to be certain.
Chat is good at a different job. It is strong at narrowing options, translating fuzzy intent into candidate products, and helping someone get unstuck when they do not know what to buy. But once a shopper is close to spending money, the conversational frame starts adding friction instead of removing it. Messages stack vertically. Product comparison becomes awkward. Variant selection is easier to miss. Side-by-side evaluation is weaker. The user also has to trust that the assistant has surfaced the right option without the usual retail signals doing as much visible work.
Even small missing details matter. OpenAI's help documentation for Instant Checkout notes that some customized items still require follow-up by email because checkout does not collect all personalization fields. That sounds like an edge case, but it reveals the broader issue: chat checkout compresses complex purchase flows into a format that works best when the product is simple and the buyer is already mostly decided.
Walmart just paid for an expensive lesson in channel fit
The deeper read is not that AI commerce is fake. It is that channel fit still rules.
Walmart was not testing an obscure plugin on a niche storefront. It was testing in-chat checkout with one of the most recognized retailers in America, plugged into the highest-profile consumer AI product on the market. If that setup still converts at a fraction of the core site, then conversational commerce is not a universal replacement for conventional ecommerce flows. It is an overlay with a narrower best-use case than the pitch suggested.
That should reset expectations for every retailer and software vendor rushing to add "buy in chat" to the roadmap. Discovery and conversion are not the same moment. A conversational interface can be excellent at the first and mediocre at the second. Teams that blur those stages together risk measuring enthusiasm at the top of the funnel and mistaking it for purchase intent at the bottom.
The Walmart result also hints at an incentive mismatch in AI checkout products. Platform teams want to keep the user inside the assistant because that proves the assistant is becoming a destination. Merchants want the highest-converting path, even if that path exits the chat window the moment the shopper is ready. Those are not automatically aligned goals.
Where in-chat buying still fits
None of this means Instant Checkout has no future. It likely works best in tighter situations: replenishment orders, low-consideration items, simple baskets, or cases where the user explicitly starts in ChatGPT and values speed over browsing depth. It may also work better when the assistant handles the pre-purchase work and then hands off to a merchant page for the last mile rather than insisting on owning the final transaction itself.
That is a less glamorous vision than "commerce moves into chat," but it is a more defensible one. AI can compress research, filter overwhelming catalogs, and tee up a shortlist faster than a standard search interface. That is real value. The mistake is assuming the same interface should also dominate the final conversion step.
Walmart's one-third conversion result is the first concrete reminder that retail gravity still exists. People will happily use AI to decide what to buy. They are much less eager to abandon familiar, high-confidence purchase environments when it is time to actually spend money. The verdict is blunt: conversational commerce belongs upstream of checkout more often than it replaces checkout, and this partnership just supplied the number that cuts through the hype.