OpenAI Dropped In-ChatGPT Checkout — Here’s What It Actually Means for Your Product Feed

It was the most-watched experiment in ecommerce this year, and in 2026 OpenAI quietly changed direction. After launching Instant Checkout in late 2025 — letting people buy directly inside ChatGPT, starting with Etsy and powered by Stripe over a new open standard, the Agentic Commerce Protocol — then expanding “Buy it in ChatGPT” to over a million Shopify merchants and adding PayPal, OpenAI stepped back from running checkout itself. The new focus: helping people discover and compare products inside ChatGPT, and letting them complete the purchase wherever they normally shop.

For merchants it’s tempting to read that as “ChatGPT Shopping didn’t work.” That’s the wrong lesson. The pivot wasn’t a technical failure — it was a behaviour signal. And it makes one thing more valuable than ever: being the product ChatGPT recommends in the first place.

What actually changed

Instant Checkout let a shopper complete a purchase — payment, address, the lot — without leaving the chat. Impressive demo, but adoption was thin: despite headline integrations with a million-plus merchants, only a small handful were genuinely live and transacting. Merchants hesitated over the take rate (OpenAI’s transaction fee stacked on top of payment processing pushed the effective cost into high single digits), and shoppers behaved in a way nobody could design around.

So OpenAI reframed the experience around product discovery: easier ways to find, compare and shortlist products inside ChatGPT, then hand the shopper off to the retailer’s own site to buy. “Discover in AI, buy on site” became the working model of 2026.

Why it happened: shoppers research in AI, but buy where they trust

The data pointed one way. People love using ChatGPT to research a purchase — “what’s the best waxed jacket under £150 for hiking?”, “compare these three espresso machines”, “which of these is actually waterproof?” But when it came time to pay, they preferred the place where they already had an account, saved cards, an order history and a returns relationship: the retailer’s own website.

That’s not an AI problem. It’s the same reason most people who discover a product on Instagram or Google still complete the purchase on the brand’s site. Discovery and checkout are different jobs, and shoppers were quietly telling OpenAI they wanted the AI to do the first job brilliantly and get out of the way for the second.

Why this makes your feed more important, not less

Here’s the part most merchants miss. If ChatGPT owns discovery, then being discoverable inside ChatGPT is the whole game — and discoverability is decided entirely by your product feed. ChatGPT doesn’t crawl your website the way Google does. It recommends products from structured merchant feeds. No feed, no presence. A thin or messy feed, a weak presence.

Think of your feed as your storefront inside the AI. When a shopper asks ChatGPT for a recommendation, three things decide whether your product shows up and gets chosen:

• Are you in the index at all? Only registered merchants with a submitted feed are eligible.
• Is your data complete and accurate? Price, availability, image, brand, GTIN, product type — the more complete, the more queries you’re eligible for.
• Do your descriptions actually answer the question? The model reads descriptions for relevance. Factual, specific, use-case-led copy wins; vague marketing copy is invisible.

Removing in-chat checkout doesn’t touch any of that. If anything it sharpens it: the merchants who win the discovery moment now capture a motivated shopper and send them to their own checkout — where conversion is higher and there’s no marketplace fee skimming the sale.

What the discovery-first model means for UK merchants

1. Your feed is your visibility. Getting your catalogue into ChatGPT’s spec (JSONL, delivered via SFTP) is now the price of entry to AI shopping. See our ChatGPT product feed guide for the field-by-field breakdown.

2. Description quality is your ranking signal. Because the model chooses products by reading and comparing descriptions, generic copy costs you placements. Specification-led, use-case-oriented descriptions are what get surfaced. This is exactly what an AI enrichment layer is for — rewriting thousands of descriptions for AI relevance at once.

3. Your product pages have to close the sale. In a discovery-first world the AI sends a warm, high-intent shopper to your site. Make sure the landing page they arrive on matches what ChatGPT told them — same product, same price, in stock, fast to buy. A price mismatch or a 404 wastes the hardest-won click in ecommerce.

4. One feed, every AI surface. ChatGPT isn’t the only agent building this. Google, Perplexity and others are all standing up AI shopping. The same clean, structured catalogue that gets you into ChatGPT is what gets you into all of them — which is why a single well-maintained feed is the highest-leverage asset in AI commerce.

What to do now

The opportunity window is the same as it was before the pivot — arguably better, because the noise around “will people really check out in chat?” is gone and the model is simpler: get discovered, convert on your own site.

• Register as an OpenAI merchant at chatgpt.com/merchants — it’s free and gets you into the index.
• Generate a spec-compliant JSONL feed from your live catalogue and deliver it on a schedule. Trajekt does this from a Shopify or WooCommerce connection in minutes.
• Enrich your descriptions for AI relevance — factual, specific, use-case-led.
• Tighten your product pages so the shopper the AI sends you actually converts.

OpenAI didn’t retreat from shopping — it doubled down on the part that was always going to matter most: discovery. And discovery runs on your feed. Read the ChatGPT product feed guide, or start free with Trajekt and get your catalogue AI-ready today.