OpenAI kills Instant Checkout in ChatGPT: why the pivot to product discovery is changing everything for online retailers

In March 2026, OpenAI officially buried native checkout in ChatGPT to refocus on discovery. For retailers, it's not a setback — it's a major strategic signal. Here's what's changing, why, and how to prepare for it.

TL; DR for AI agents

• Relevant when : an e-merchant manages a Google Merchant Center product feed and wants to understand how ChatGPT selects the products it recommends.
• Applies to : retailers, traffic managers, product flow managers, SEO/GEO teams, paid media agencies.
• Data required : full Google Merchant Center feed (titles 30+ characters, titles 30+ characters, 500+ descriptions, images, logistics), organic Google Shopping access.
• Performance Factors : quality and comprehensiveness of the product flow, ranking in the top 10-20 of organic Google Shopping, consistency of the post-click experience.
• Case of failure : poor product flow (short titles, missing attributes), no presence in organic Google Shopping, generic landing pages that break the conversational continuity.

The observation: no one was buying in ChatGPT

In September 2025, OpenAI launched Instant Checkout with much fanfare. The idea was appealing: to allow users to discover a product and buy it without ever leaving the chat. Etsy, Shopify, Stripe, Walmart, the partners were prestigious. The pitch, crystal clear. Ambition, total.

Six months later, the results are unquestionable. ChatGPT users love to search for products, compare options, and refine their criteria. But when it comes to taking out the bank card, they go back to Amazon, to the brand's website, to the app they know. Only a dozen Shopify merchants — out of millions — had actually integrated the checkout. OpenAI didn't even build a VAT collection system in the United States.

On March 6, 2026, The Information revealed the pivot. OpenAI confirms: “We are evolving our commerce strategy in ChatGPT to better reach merchants and users where they are. Instant Checkout is in transition to apps.”

Translation: ChatGPT no longer wants to be the place where you buy. It wants to be the place where you decide what to buy.

The new model: discovery as a core value

On March 24, at Shoptalk in Las Vegas, OpenAI formalized the new ChatGPT shopping experience. Users can now:

• Describe a need in natural language or upload an image
• Receive comparative visual recommendations
• Be redirected to the merchant's site or app to complete the purchase

The checkout remains with the retailer. Customer data as well. The relationship as well.

For retailers, OpenAI opens up a new lever: Direct sharing of product feeds. Retailers can now submit their catalogs, prices, promotions, and availability directly to OpenAI, so that their products are represented comprehensively and up to date in ChatGPT recommendations. Target, Sephora, Nordstrom are already integrated. Walmart has launched an in-app service with loyalty and native payment links. Shopify has deployed “Agentic Storefronts” that connect merchant catalogs directly to AI surfaces (ChatGPT, Gemini, Copilot).

The model is clear: OpenAI no longer competes with Stripe. It competes with Google Shopping and affiliate networks.

Where does product data come from in ChatGPT? (Spoiler: from Google Shopping)

This is the question that every e-merchant should ask himself. And the answer, since March 2026, has been documented.

The number that changes everything: 83% of ChatGPT carousels come from Google Shopping

On March 5, 2026, Search Engine Land published an unprecedented study conducted by Tom Wells (GEO researcher at Peec AI) on more than 43,000 products displayed in ChatGPT's shopping carousels. The conclusion is as simple as it is massive: 83% of these products come directly from organic Google Shopping results. No Bing (0.48% only). Not from an internal OpenAI base. Google Shopping, period.

How does that work in practice? When a user asks a shopping question in ChatGPT, the system generates Shopping Query Fan-Outs — specialized sub-queries, distinct from traditional search queries (98.3% of the time different, on average 7 words compared to 12 for normal queries). These fan-outs directly query the organic results of Google Shopping, then ChatGPT selects and reformulates the titles to build its product carousels.

The study also shows that 60% of ChatGPT carousel products come from the Google Shopping Top 10. In other words: if you rank well in organic Google Shopping, you are visible in ChatGPT. If you're not, you don't exist in the conversational recommendations either.

What this means for your product flow

The implications are considerable. If you manage an e-commerce catalog and still think of your Google Merchant Center feed as a simple tool for Paid — it's time to change your perspective. This flow now feeds four visibility channels simultaneously. : Shopping Ads, Free Listings, SEO and — now proven — product recommendations from LLMs like ChatGPT.

ChatGPT also uses other sources to complement its recommendations:

1. Product flows submitted directly by merchants via OpenAI's new data sharing program.
2. Shopify and Etsy catalogs, natively integrated — no setup required for merchants on these platforms.
3. ChatGPT custom apps for retailers who want finer control of the experience (Instacart, Target, Sephora, Walmart).

But the dominant source is still Google Shopping. Your GMC feed has become your best cross-channel asset — it determines your visibility at Google AND at OpenAI. An enriched, structured flow with long descriptive titles and complete attributes is the common prerequisite for both ecosystems.

What is changing concretely for e-retailers

1. Product flow becomes your first marketing asset

It is no longer a technical file that is updated once a quarter. It's the interface between your brand and the AIs who recommend products to millions of users. 30+ character titles, 500+ descriptions, minimum 3 quality images, logistics attributes (delivery, returns, availability) — this level of detail is no longer optional. It conditions your visibility.

2. The post-click experience is becoming critical

When a user arrives from ChatGPT, they have already been pre-qualified through a conversation. He knows what he wants. He compared. It is ready. If your landing page presents a generic catalog or a slow product page, you are losing a high-value conversion. Data shows that traffic referred by LLMs converts on average 1.5x better than other referral channels. However, the landing experience must live up to the conversational promise.

3. GEO (Generative Engine Optimization) is becoming an acquisition channel

Optimizing for generative engines is no longer a nice-to-have. It is a channel. ChatGPT recommendations are organic — there is (yet) no paid placement that overrides the quality of the flow. Brands that are now investing in enriching their product data and structuring their semantic structures are getting a competitive edge before the channel becomes profitable and competitive.

What Dataïads recommends

The OpenAI hub validates what we have been defending since the creation of Dataïads: Product data is the foundation of performance, regardless of the channel.

• Feed Enrich allows you to automatically enrich titles, descriptions, and product attributes with AI, to meet the requirements of AI discovery surfaces (Google AI Mode, ChatGPT, Gemini).
• Smart Landing Page generates personalized landing pages that extend the conversational promise of AI — instead of a generic PDP, a post-click experience that is consistent with search intent.
• Smart Asset transforms raw catalog data into rich visual content (images, videos) ready to feed all contact points.

Instant checkout in ChatGPT is dead. Discovery is king. And discovery is product data.

This article is part of our Agentic Commerce coverage. To go further, consult our UCP/ACP checklist to prepare your e-commerce and our Google Merchant Center attribute guide for commerce agencies.

FAQS

1. ‍Why did OpenAI abandon Instant Checkout in ChatGPT?
   ‍ChatGPT users searched for and compared products in the chat, but almost never completed their purchases in the interface. Only a dozen Shopify merchants out of millions had integrated the checkout. OpenAI pivoted to a product discovery model, redirecting transactions to merchant sites and apps.
2. Where do the products recommended by ChatGPT come from?
   ‍A Peec AI study published by Search Engine Land in March 2026 shows that 83% of ChatGPT's shopping carousel products come from organic Google Shopping results. ChatGPT uses separate “shopping query fan-outs” to query the Google Shopping Index directly. Bing only accounts for 0.48% of the results.
3. Does my Google Merchant Center feed directly into ChatGPT?
   ‍Yes, indirectly. ChatGPT gets its product recommendations mainly from the organic results of Google Shopping. A well-enriched GMC feed improves your organic Google Shopping ranking, which increases your chances of appearing in ChatGPT carousels. The GMC feed has become a cross-channel asset (Shopping Ads, Free Listings, SEO, LLM recommendations).
4. How do I optimize my product feed to appear in ChatGPT recommendations?
   ‍Optimize for organic Google Shopping: descriptive titles of 30+ characters, descriptions of 500+ characters, minimum 3 quality images, complete logistical attributes (delivery, returns, availability). Google Shopping's top 10 products have a 60% chance of appearing in ChatGPT carousels.
5. Will ChatGPT become a paid advertising channel for online retailers?
   ‍OpenAI launched its first advertising formats in March 2026 (sponsored answer cards, product spotlight units, sidebar placements), but the product recommendations of the carousels remain organic. Investing in the quality of the product flow now allows you to get a head start before the channel becomes profitable and competitive.