ChatGPT Shopping: How Products Get Into ChatGPT (2026)

ChatGPT shopping is the product search experience inside ChatGPT: you describe what you want, or upload a photo, and ChatGPT returns visual product results with prices, merchants and comparisons. Products get there two ways: ChatGPT's search crawler reads public product pages, and approved merchants send OpenAI a structured product feed. As of September 2026, most purchases finish on the merchant's own site or app, because OpenAI stepped back from in-chat Instant Checkout in March 2026.

This guide explains how products appear, what the feed requires, what happened to Instant Checkout, and what a merchant can do this month.

How does ChatGPT shopping work?

When a question looks like a shopping question ("quiet cordless vacuum under $300 for a small flat"), ChatGPT searches for products, shows them as visual results, and writes a comparison based on the criteria in the question.

OpenAI rebuilt this experience in March 2026. According to CNBC's report on the update, shoppers can find products by uploading images or describing items, including budget, preferences and other constraints, and ChatGPT shows more visual results to compare offers. OpenAI told CNBC it had improved speed, relevance and product coverage.

The product data comes from two places:

1. Crawled web pages. ChatGPT search uses OAI-SearchBot. OpenAI's crawler documentation says sites that opt out of OAI-SearchBot won't be shown in ChatGPT search answers (though they can still appear as navigational links).
2. Merchant product feeds. Merchants approved by OpenAI push a structured catalog. OpenAI's commerce getting-started guide says feeds give ChatGPT the catalog data it needs to index products and present accurate information in shopping experiences.

OpenAI hasn't published how it ranks products against each other. The feed spec does say two useful things: recommended attributes such as media and reviews help relevance, and being eligible for search does not guarantee a product is displayed.

What happened to ChatGPT Instant Checkout?

Instant Checkout was OpenAI's attempt to let people buy without leaving the chat.

• September 2025: OpenAI launched Instant Checkout in the US, starting with Etsy sellers, with Shopify merchants to follow. It ran on the new Agentic Commerce Protocol, built with Stripe.
• March 2026: OpenAI reversed course. Agentic Commerce Feed's write-up dates the change to March 6 and quotes OpenAI saying it would prioritize ChatGPT search and product discovery, with ACP as the infrastructure connecting users to merchants.
• March 24, 2026: OpenAI announced the revamped shopping experience. In CNBC's report, OpenAI said the first version of Instant Checkout did not offer the flexibility it wanted, so it was letting merchants use their own checkout experiences while it focused on product discovery.

What replaced it, per the same CNBC report:

• Feeds and promotions. Merchants can share product feeds and promotions with OpenAI so their products are fully represented. CNBC named Target, Sephora and Nordstrom as already supporting the new discovery experience.
• Merchant apps in ChatGPT. Retailers can build their own apps inside ChatGPT, which gives them control of the experience and the transaction. Walmart launched one supporting account linking, loyalty and Walmart payments.
• Shopify's route. Shopify said its merchants can connect storefronts to its catalog and complete purchases through an in-app browser, and announced an "Agentic Plan" for merchants without a Shopify storefront.

One nuance: OpenAI's developer docs still publish the Agentic Checkout Spec and Delegated Payment Spec, and the product feed spec treats checkout as a separately enabled integration that is off by default. So in-chat checkout technology still exists, but for most merchants in 2026 the job is discovery: be found, be described correctly, and send the shopper to a checkout you control.

What is the ChatGPT product feed?

The product feed is a structured catalog file (or API stream) that you send to OpenAI. It follows the product feed part of the Agentic Commerce Protocol. ACP is described on its official site as an open standard for commerce flows between buyers, AI agents and businesses, developed by OpenAI and Stripe under the Apache 2.0 license.

Who can submit a feed?

OpenAI's getting-started guide says onboarding product feeds is currently available to approved partners, with an application form linked from the docs. If you sell through Shopify, check Shopify's ChatGPT integration first; it may be the faster route.

How is the feed delivered?

From OpenAI's file upload overview and getting-started guide (as of September 2026):

Topic	Requirement
Methods	File upload via SFTP, or an API for creating and upserting products
Recommended pattern	Full feed once a day by file, updates during the day by API
File formats	Parquet preferred (ideally zstd compression); jsonl.gz, csv.gz, tsv.gz also supported
Encoding	UTF-8
Cadence	Full snapshot at least daily
Promotions	API only
Removed products	Kept for up to 14 days after they disappear from snapshots

Which fields are required?

The products spec lists nine required fields for a discovery feed, one row per purchasable item or variant:

Field	What it holds
item_id	Stable, unique ID per item or variant; never reused
title	Product name including variant (aim for 150 characters or fewer)
description	Factual plain-text description (up to about 5,000 characters)
url	Product page, with the variant selected where possible
brand	Brand as shown on the product page
seller_name	Name of the seller supplying the offer
image_url	Direct URL to the main image for this variant
availability	in_stock, out_of_stock, pre_order, backorder or unknown
price	Regular price with currency, e.g. 79.99 USD

A minimal record in JSONL looks like this:

{"item_id":"MUG-350-BLUE","title":"Blue ceramic mug, 350 mL","description":"Dishwasher-safe glazed ceramic mug with a handle.","url":"https://example.com/products/mug-blue","brand":"Northline","seller_name":"Northline Home","image_url":"https://example.com/images/mug-blue.jpg","price":"18.00 USD","availability":"in_stock"}

Useful optional fields include sale_price, gtin and mpn identifiers, variant attributes, shipping, returns, and review_count with star_rating. The flag is_eligible_search defaults to true; set it to false to pull a product from search. The spec also describes a Google-compatible feed format, so a clean Google Merchant Center feed is a good head start.

OpenAI's best practices page adds practical rules: factual descriptions, valid and encoded URLs, variant-specific titles and images, and UTM parameters on feed URLs if you want to track feed clicks.

What can't be sold?

The getting-started guide lists prohibited categories, including adult content, age-restricted products such as alcohol, nicotine and gambling, weapons, prescription-only medicines and unlicensed financial products. OpenAI can remove products or sellers that break these rules.

Does Product schema matter for ChatGPT shopping?

Product structured data is schema.org markup (Product, Offer, AggregateRating) that states a product's name, price, availability and ratings in machine-readable form. Google documents it for merchant listings and product snippets.

OpenAI has not said whether ChatGPT reads Product schema. Here is the reasoning for adding it anyway, labeled as reasoning:

• OAI-SearchBot crawls your product pages. Markup that states price and availability plainly reduces the chance of a crawler misreading a page full of related products and crossed-out prices.
• If you also send a feed, the page and the feed should say the same thing. Product markup generated from the same data source keeps them consistent.
• It already pays off in Google Shopping surfaces, so the work isn't wasted.

A compact example:

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Blue ceramic mug, 350 mL",
  "brand": { "@type": "Brand", "name": "Northline" },
  "sku": "MUG-350-BLUE",
  "image": "https://example.com/images/mug-blue.jpg",
  "offers": {
    "@type": "Offer",
    "price": "18.00",
    "priceCurrency": "USD",
    "availability": "https://schema.org/InStock",
    "url": "https://example.com/products/mug-blue"
  }
}

Our schema markup generator writes Product JSON-LD, and the schema validator checks what is live on a page.

What merchants can do now

A practical order of work, from cheapest to most involved:

1. Allow OAI-SearchBot. Check robots.txt and your CDN or firewall rules. Our AI crawler checker shows which AI bots are allowed. GPTBot (training) is a separate setting; you can block it and still allow OAI-SearchBot.
2. Make product pages readable without JavaScript. Price, availability, variant names and key specs should be in the server-rendered HTML. Check with the AI crawler view.
3. Write descriptions that answer shopping questions. ChatGPT matches products against criteria like "under $300", "for a small flat", "dishwasher-safe". State dimensions, materials, compatibility and use cases as plain facts.
4. Add Product markup that matches the visible page.
5. Clean your feed. Stable IDs, real brand and seller names, direct image URLs, explicit availability, GTINs where assigned. A tidy Google Merchant Center feed covers most of this.
6. Apply for feed onboarding through the form linked in OpenAI's commerce docs, or use Shopify's integration if you are on Shopify.
7. Earn third-party mentions. For "best X" questions, ChatGPT often draws on reviews, comparison articles and forums. Being in those sources matters as much as your own page. See our guide to ChatGPT SEO.
8. Consider an app only if you are large. Merchant apps in ChatGPT give control of checkout and loyalty, but they are a development project. For most stores, feed plus a good product page is the right scope.

How to track whether ChatGPT recommends your products

Run the shopping questions your customers ask, in ChatGPT, and record which products and merchants appear. Do it for a fixed prompt set over several weeks, because answers vary between runs.

At brand level, aeotime (our product) runs your prompts in ChatGPT and seven other AI engines and reports whether your brand is mentioned, its position, the sources cited and competitors named. We collect answers through official APIs with web search on, which can differ slightly from the logged-in ChatGPT app, including in how product results are displayed. Product-level listing audits for AI shopping and marketplaces such as Amazon, Trendyol and eBay are coming soon; see marketplace optimization. Other vendors, including Profound (Enterprise plan) and Peec AI, offer shopping analytics too; our AI visibility tools roundup compares them.

ChatGPT shopping checklist

• OAI-SearchBot allowed in robots.txt and not blocked at the firewall
• Price, stock and key specs visible in raw HTML on every product page
• Product JSON-LD that matches the page
• Feed with the nine required fields, refreshed at least daily
• Applied for OpenAI feed onboarding, or connected through Shopify
• A fixed set of shopping prompts checked weekly in ChatGPT

Start with the first two items. They cost nothing and decide whether ChatGPT can see your products at all. Run a free check with the AI visibility checker to see how ChatGPT describes your brand today.