The Brands Are Taking Back Control
OpenAI killed Instant Checkout. Walmart said the experience was “unsatisfying.” Now every major retailer is building their own AI — and the implications for commerce are enormous.
In the last two weeks, something quietly historic happened in e-commerce. OpenAI officially stepped away from Instant Checkout — the feature that was supposed to let people buy products without leaving ChatGPT. And then Walmart, one of the original launch partners, went public with exactly why: conversion rates for purchases made inside ChatGPT were three times worse than when users just clicked out to Walmart.com.
Daniel Danker, Walmart's EVP of AI Acceleration, called the experience “unsatisfying.” Customers feared that “when checkout happens automatically after every single item, they're going to receive five boxes when they actually just want it all in one.”
This isn't just one company adjusting a feature. It's the opening act of a much bigger story — one where the world's largest brands realize that outsourcing the customer relationship to a third-party AI is a strategic mistake.
The Rise and Fall of Instant Checkout
When OpenAI launched Instant Checkout on September 29, 2025, the pitch was compelling. Buy anything without leaving ChatGPT. Powered by the Agentic Commerce Protocol (ACP), co-developed with Stripe. Merchants could enable it with a single line of code. Over a million Shopify merchants — Glossier, SKIMS, Spanx, Vuori — were slated to join. OpenAI charged a 4% transaction fee on every completed purchase.
On paper, it looked like the future. In practice, it had problems that should have been predictable.
The problems stacked up fast:
Limited selection. Six months in, only about 200,000 Walmart products were available — and pricing and inventory data was often stale.
Single-item only. No multi-item carts. You couldn't bundle a laptop with a case and a mouse. Every item was a separate transaction.
No sales tax. OpenAI hadn't solved state-by-state tax collection and remittance — a problem retailers have spent decades building infrastructure for.
Unfamiliar payment flow. Customers hesitated. The checkout experience felt nothing like what they were used to on retailer sites.
High-consideration items excluded. Televisions, furniture, appliances — the products people actually need help choosing — were off the table because the chat interface couldn't support them.
The fatal issue wasn't any single flaw. It was the fundamental assumption: that people who research products in a conversational AI would also want to buy there. They didn't. They used ChatGPT to figure out what to buy, then went to the retailer to actually buy it.
Walmart's 180: From Checkout to Sparky
Here's what makes the Walmart story so instructive. They didn't just walk away from Instant Checkout — they replaced it with something that reveals a completely different philosophy.
Starting the week of March 25, 2026, Walmart is embedding Sparky — its own AI shopping assistant — directly inside ChatGPT. Instead of OpenAI handling the transaction, Walmart's AI handles the entire experience. Carts sync across Walmart.com, the mobile app, and ChatGPT. Walmart retains customer data control. OpenAI doesn't process transactions.
The early numbers are striking. Sparky inside ChatGPT is converting at roughly 70% of Walmart.com's rate — a massive improvement over Instant Checkout's one-third rate. ChatGPT is sending Walmart approximately 2x more new customers than search engines. And Walmart negotiated from a position of strength: they pay a flat API fee, not a revenue share.
Danker described the approach as bringing “the Walmart store to the customer” rather than forcing them into a disconnected experience.
This matters because Sparky wasn't an afterthought. Walmart launched it on June 6, 2025 on its own properties first, and the traction has been remarkable:
of Walmart app users have interacted with Sparky at least once
higher average order value for customers who engage with Sparky
in Walmart Connect retail media revenue for FY2026, up 46% year-over-year
The Pattern: Every Giant Is Saying the Same Thing
Walmart isn't an isolated case. Look at what's happening across the industry.
Amazon built Rufus — and walled off its garden
Amazon's AI assistant launched in beta in February 2024 and has since been used by 250 million shoppers. Monthly active users grew 140% year-over-year. The revenue impact? An estimated $10 billion in annualized incremental sales. Customers who engage with Rufus are 60% more likely to complete a purchase.
And Amazon didn't just build its own AI — it actively blocked everyone else. They've shut out 47 AI bots from their platform, including agents from OpenAI, Google, Anthropic, Perplexity, and Microsoft. In March 2026, a federal court granted Amazon a preliminary injunction blocking Perplexity AI's Comet browser from accessing Amazon. The message is clear: if you want to shop on Amazon, you use Amazon's AI.
Nike launched NikeAI
In August 2025, Nike rolled out NikeAI to all U.S. iOS users. Built on foundation models fine-tuned by domain experts who understand Nike's brand DNA — handling natural-language queries for running shoes, team gear, specific colorways, and filters not available in the standard UI. As Nike CTO Muge Erdirik Dogan put it: “It's a meaningful shift in how we connect our athletes with products that fit their needs, moments, and goals.”
Target owns its own ChatGPT app
Target launched a dedicated ChatGPT app for the 2025 holiday season with personalized recommendations, multi-item baskets, fresh food products, and checkout via Drive Up, Order Pickup, or shipping. Traffic from ChatGPT to Target has been growing 40% month-over-month. And in February 2026, Target became one of the first retailers to test contextual advertising inside ChatGPT — ads from Target and its Roundel retail media partners appearing alongside shopping conversations.
Shopify and Etsy are pulling back too
Shopify is ending ChatGPT native checkout, routing transactions back to merchant storefronts. Etsy is building a dedicated ChatGPT app to maintain control of the customer experience. The merchants themselves are saying: we need to own this.
The Amazon Playbook, Replayed
There's a historical parallel here that's almost too perfect.
In the early days of e-commerce, Amazon offered brands incredible distribution. List your products on Amazon and reach millions of shoppers. The deal seemed too good to refuse. Brands flooded in. And then, gradually, they realized what they'd traded away.
They lost control of the customer relationship. They lost access to customer data. They lost the ability to build brand loyalty — because to the customer, the relationship was with Amazon, not the brand. They became interchangeable inventory inside someone else's interface.
Now ChatGPT is following the same script. Incredible distribution — over 400 million users. A conversational interface that could surface your products to people with genuine buying intent. But if the AI handles the recommendation, the checkout, and the transaction, what's left for the brand?
The smartest brands figured out the Amazon lesson a decade ago: distribution without relationship is a trap. Now they're applying that lesson to AI before it's too late.
Why Third-Party AI Creates a Subpar Experience
The brands aren't just protecting their business interests. They're also right about the experience.
A generic AI shopping through an API can tell you product specs and prices. What it can't do is understand a brand's world the way the brand does. Consider what you lose:
Brand context disappears.
Nike doesn't just sell shoes — it sells to runners training for their first marathon, to parents buying cleats for their kid's season, to collectors who care about colorway drops. A generic AI strips all of that context. It reduces Nike to SKUs and prices.
Product knowledge is shallow.
An AI reading product descriptions can tell you the weight of a laptop. It can't tell you that this particular model runs hot under sustained workloads, or that the keyboard is widely considered the best in its class, or that the battery life drops significantly if you use it with an external monitor. Brand-owned AI can be trained on deep product knowledge that no third party has.
The purchase journey breaks.
Shopping isn't just picking a product and paying for it. It's checking if your size is in stock, applying loyalty points, choosing between shipping and in-store pickup, adding gift wrapping, using a promo code. Instant Checkout couldn't handle any of this. A brand's own AI can integrate with the full checkout stack.
Trust is harder to build.
When a customer asks “will this fit me?” inside ChatGPT, they're trusting OpenAI with a question that really belongs to the retailer. A brand-owned AI can leverage return data, fit models, customer reviews, and sizing algorithms that no external AI has access to.
The Numbers Tell the Story
The data from Q4 2025 and early 2026 makes the case clearly:
Amazon Rufus
250M users. 60% higher purchase completion. $10B in estimated incremental annual revenue. Projected $1.2B in profit contribution by 2027.
Walmart Sparky
50% of app users engaged. 35% higher average order value. Converting at 70% of Walmart.com rates inside ChatGPT (vs. 33% for Instant Checkout).
Target ChatGPT App
40% month-over-month traffic growth. First retailer testing contextual ads inside ChatGPT.
OpenAI Instant Checkout
3x worse conversion than retailer sites. Single-item only. No tax support. Phased out after 6 months.
What This Means for the Rest of the Market
Amazon, Walmart, Nike, Target — these companies can afford to build their own AI shopping experiences from scratch. They have the engineering teams, the data, the budgets. But what about everyone else?
McKinsey estimates agentic commerce could touch $1 trillion in U.S. retail revenue by 2030. Morgan Stanley projects nearly 50% of American shoppers will use AI agents by then, representing $115 billion in additional e-commerce spending. This isn't a small market and it isn't optional.
For the thousands of mid-market and D2C brands that don't have Amazon's resources, the lesson from the past few weeks is clear but the path forward is harder. You can't outsource AI shopping to ChatGPT — it creates a subpar experience and you lose the customer relationship. But you also can't spend $100 million building your own Rufus.
This is the gap that matters most right now. Not the gap between humans and AI, but the gap between the handful of companies that can build brand-owned AI commerce and the thousands that can't.
What Comes Next
OpenAI is pivoting to an app-based model — letting retailers build dedicated ChatGPT apps while OpenAI focuses on search and discovery. The Agentic Commerce Protocol lives on as an open standard. The platform play isn't dead; it's just being redrawn.
But the fundamental insight from Walmart, Amazon, Nike, and Target won't change: the brands that own their AI experience will win. The ones that outsource it to a platform will end up exactly where they ended up with Amazon — as interchangeable commodity inventory inside someone else's interface.
The giants are starting to figure this out. Curious to see how the rest of the market reacts over the next few months.
Because right now, the biggest shift in e-commerce since the smartphone isn't about whether AI will change shopping. It's about who controls the AI that does.
We're building the infrastructure that lets every brand own their AI shopping experience — not just the giants. If that resonates, we'd love to talk.