Walmart just killed ChatGPT checkout — it converted 3x worse

Walmart tested letting people buy products directly inside ChatGPT for five months — and the results were brutal. Purchases completed inside the chat converted at one-third the rate of transactions where shoppers clicked through to Walmart's website. That's a 66% drop in sales performance, and it was enough for Walmart to pull the plug.

Daniel Danker, Walmart's Executive Vice President of Product and Design, called the in-chat buying experience "unsatisfying." Starting the week of March 25, Walmart is replacing it with something completely different.

What went wrong with buying inside ChatGPT

Last November, Walmart made about 200,000 products available through OpenAI's Instant Checkout — a feature that let you find, select, and pay for items without ever leaving the ChatGPT window. On paper, it sounded like the future of shopping.

In practice, shoppers hated it. The biggest problem? Every product recommendation triggered its own separate transaction. If ChatGPT suggested five items, that meant five individual orders — five boxes arriving at your door on five different days. Customers intuitively rejected this.

There were also technical hurdles OpenAI never fully solved: real-time inventory sync across millions of products, U.S. state sales tax collection, and fraud prevention when AI agents handle payments.

The pivot: Sparky moves into ChatGPT

Instead of letting ChatGPT handle the sale, Walmart is now embedding Sparky — its own AI shopping assistant — directly inside ChatGPT. Here's what changes:

The logic is simple: let AI do what it's good at (finding products, answering questions, making recommendations) and let Walmart's own system do what it's good at (closing the sale).

The numbers that matter for businesses

Here's what makes this story more nuanced than a simple failure:

ChatGPT is bringing Walmart roughly twice the rate of new customers compared to search engines. The people discovering Walmart through ChatGPT are a different demographic than typical Walmart shoppers — younger, more affluent, and more willing to try new products.

Users who interact with Sparky show 35% higher average order values than other shoppers. And the top-selling categories through AI are telling: vitamin and protein supplements (especially from people asking about GLP-1 weight-loss drugs), plus automotive, beauty, and home improvement products.

So while checkout inside ChatGPT flopped, AI-powered product discovery is working. Walmart just learned the hard way that discovery and purchase need to happen in different places.

OpenAI abandons Instant Checkout entirely

This isn't just a Walmart problem. OpenAI is phasing out Instant Checkout across its entire platform, shifting to what it calls "apps" — where purchases happen inside connected merchant services rather than natively in ChatGPT.

An OpenAI spokesperson confirmed the shift: "We are evolving our commerce strategy within ChatGPT to better meet merchants and users where they are. Instant checkout is transitioning to apps, where purchases can occur more seamlessly."

Translation: the dream of AI handling your entire shopping trip — from browsing to buying — isn't ready. Not because the AI can't recommend good products, but because people don't trust a chatbot with their credit card the way they trust a retailer's own checkout page.

What this means if you sell things online

If you're a marketer, e-commerce manager, or small business owner, here's the takeaway:

The era of AI shopping isn't dead. It's just being rebuilt — with the retailer, not the AI platform, in control of the sale.

Related Content — Get Started with Easy Claude Code | Free Learning Guides | More AI News