OpenAI has quietly walked back one of its most ambitious commerce bets. Instant Checkout — the feature launched in September 2025 that let users buy products directly inside ChatGPT — has been retired. In its place: the Agentic Commerce Protocol (ACP), a discovery-first model that shifts ownership of the checkout experience back to the retailers themselves.

The swap happened on March 24, 2026. And it reveals something important about how AI is actually integrating into commerce — and what that means for any business that sells things.

What Instant Checkout Was, and Why It Flopped

When OpenAI launched Instant Checkout, the pitch was compelling: imagine asking ChatGPT “buy me running shoes under $150 for wide feet” and having the purchase completed inside the conversation. No redirects, no new tabs, no hunting through product pages.

In practice, it never got traction. By the time it was discontinued, only around 30 Shopify merchants had completed the onboarding and fewer than 200,000 Walmart products were available through the feature. The friction was real on both sides. Merchants were reluctant to hand over checkout to a third party — losing control of loyalty programs, payment data, and post-purchase experience. Customers found the product selection too thin to bother.

The core problem: OpenAI was asking retailers to give up the most valuable part of the transaction. Nobody wanted that.

What Replaced It: The Agentic Commerce Protocol

The Agentic Commerce Protocol is a fundamentally different model. Instead of OpenAI owning checkout, ACP lets merchants share structured product feeds, promotions, and catalog data so that their inventory surfaces in ChatGPT’s product discovery and search results. When a user finds something they want to buy, they’re handed off to the retailer’s own environment to complete the purchase.

Retailers keep their checkout, their loyalty programs, and their customer data. ChatGPT becomes the discovery layer, not the store.

The eight retailers in the first wave of ACP are a serious list: Walmart, Target, Sephora, Nordstrom, Lowe’s, Best Buy, Home Depot, and Wayfair. These aren’t pilot partners — they’re the anchors of American retail.

Walmart has built the most integrated experience so far. Through a partnership with OpenAI, Walmart’s AI commerce agent Sparky now has its own in-ChatGPT experience. A user can start with a natural language product question in ChatGPT, get Walmart-powered results, and move seamlessly into a Walmart-managed environment that supports account linking, loyalty points, and Walmart Pay. It’s discovery in ChatGPT, everything else in Walmart.

The rollout covers all ChatGPT tiers — Free, Go, Plus, and Pro — meaning this is now in front of hundreds of millions of users.

The Pattern Here Is Not Unique to Commerce

OpenAI tried to own the whole transaction. It couldn’t. So it built a protocol that lets the existing players keep what they value most (the customer relationship), while taking what it has to offer (reach, intent detection, discovery).

This is a recurring theme in how AI is actually landing in business contexts. The companies that try to disintermediate existing workflows wholesale tend to fail. The ones that slot into the workflow at the point of highest leverage — and hand off gracefully — tend to succeed.

The Agentic Commerce Protocol is also closely related to what Shopify announced the same week. Shopify’s Agentic Storefronts — which opened to all eligible US merchants on the same day as OpenAI’s ACP announcement — connects 5.6 million Shopify merchants to ChatGPT, Copilot, Gemini, and other AI surfaces through a single setup step. Shopify and OpenAI are effectively building two sides of the same system: Shopify handles the merchant feed, OpenAI handles the consumer discovery surface.

What This Means for Business

1. Product discoverability is no longer just a search engine problem.

If your products aren’t structured in a way that AI agents can understand, you’re already invisible to a growing share of buyer intent. ACP requires merchants to share machine-readable product data — not just optimized product pages. Businesses that invest in structured product data now are building an advantage that compounds as AI commerce protocols expand.

2. Checkout ownership matters more than it did six months ago.

OpenAI’s retreat from Instant Checkout validates what most enterprise retailers already knew: the checkout experience is not something you hand to a third party. The brands that build strong AI-native checkout flows within their own ecosystem will own the end-to-end customer relationship as AI commerce scales.

3. The “agent in the middle” model is here.

Whether it’s Walmart Sparky, a custom AI agent for a retailer, or an AI workflow that helps a procurement team find and approve vendors — the pattern is the same. AI sits in the middle of discovery and recommendation. Humans (or brand systems) close the loop. The businesses building those middle layers now are the ones that will define how their industry works in two years.

4. For businesses not in retail, the same logic applies.

The shift from “AI does the whole thing” to “AI handles discovery and intent, you handle the relationship” maps onto almost every professional service, B2B sales process, and high-consideration purchase. Your AI implementation strategy should reflect this — not try to automate the relationship, but use AI to surface the right opportunities and hand them to the right person.

---

OpenAI’s Instant Checkout pivot is a useful data point for any business leader thinking about AI strategy. The big tech companies are figuring this out in public, and the lesson so far is: AI works best when it augments the customer journey, not when it tries to replace it. The Agentic Commerce Protocol isn’t a failure story — it’s a course correction that makes the whole ecosystem more viable.

For a deeper walkthrough of tools like this and how they fit together, the free Working With Claude field guide covers the ecosystem end to end. Get the guide.