Why Is My Product Feed Accepted but My Products Never Appear in ChatGPT?
Because an accepted feed only makes products eligible; it does not rank them. The common causes are these. The accepted feed is a Google Merchant Center feed, which reaches ChatGPT only through Google Shopping results, or your application to OpenAI's own programme is still on its waitlist. Eligibility flags or market settings exclude the products. The questions you test do not trigger shopping results. The feed lacks the attributes buyers ask about. Your product pages contradict the feed or cannot be crawled. And rivals have stronger reviews, editorial coverage or retailer presence. When several sellers carry the same product, OpenAI has said it weighs availability, price, quality and whether the seller is the maker. Work through the causes in that order, from plumbing to persuasion.
Is ChatGPT actually reading your feed?
Check this first, because it is the most common misunderstanding:
- A Merchant Center feed is not a ChatGPT feed, but it still matters. ChatGPT does not read your Google Merchant Center account, yet most products in its shopping carousels come from Google Shopping results, which that feed powers. OpenAI runs its own feed programme, which merchants apply to join; OpenAI's merchant page says applicants go on a waitlist and that a self-serve portal is planned.
- Shopify and Etsy are the exceptions. OpenAI says Shopify and Etsy catalogues are already integrated, with no application needed. On Shopify, check that your products are eligible in the Agentic Storefronts sales channel settings.
- The Google-compatible format needs confirmation. OpenAI accepts Google-style files only after it confirms that format for your registered feed. A file uploaded in the wrong format may not be read as you expect.
If you are not in the programme, ChatGPT can still find products by crawling your site, but the feed is not what decides it.
Could settings or markets exclude your products?
Yes. OpenAI's feed specification includes controls that change eligibility:
- is_eligible_search: search eligibility is on by default, but setting it to false removes a product from search and checkout. Older aliases such as enable_search do the same.
- Markets: the standard OpenAI-format upload currently targets the US, and the shared format represents the US, Canada and Mexico. Products for other markets may not appear to shoppers elsewhere.
- Required fields: item ID, title, description, URL, brand, seller name, image URL, price and availability are required. Rows missing them, or with placeholder values, may be dropped.
- Availability: "unknown" does not assert that a product is in stock, and out-of-stock products are unlikely to be recommended.
- Shipping: OpenAI treats an omitted shipping price as unknown, not free, which can make an offer look worse than a rival's.
Check any validation feedback from OpenAI or your feed provider, and fix rows flagged as invalid.
Are you testing questions that trigger shopping results?
Often not. Shopping results appear for commercial questions, and not every product question triggers them. Profound's help documentation notes that prompts with words like "buy", "shop", "where to find", or a price or budget tend to trigger shopping, while informational questions may not. Evertune describes ChatGPT's shopping results as three to five products with prices and links, so the space is small. Test questions the way buyers ask when they are ready to purchase, such as "best waterproof hiking boots for wide feet under $200", and include some that name your product type and key attribute. Searching for your exact product name is a useful sanity check that ChatGPT knows it exists, but it is not how most buyers ask.
Does your feed describe what buyers ask for?
A feed that passes validation can still be too thin to match. Buyers ask for attributes: size, material, compatibility, use case, dietary needs, age range. If your titles say "Classic Boot – Brown" and your descriptions repeat marketing copy, the assistant has little to match against "wide-fit waterproof boots for hiking". Improve:
- Titles with product type, key attributes and variant, within 150 characters.
- Descriptions with factual detail, up to 5,000 characters, in plain text.
- Variants grouped properly with a group ID and variant attributes.
- Identifiers: real GTINs where they exist; OpenAI warns against inventing them.
- Images that show the product clearly.
How to structure catalogue data field by field is covered in how ecommerce brands can structure product data for AI recommendations.
Do your product pages agree with the feed?
ChatGPT also reads your website. OpenAI says you do not need a feed if ChatGPT already crawls your site, but a feed gives you more control. That means the two should agree. Check that:
- OAI-SearchBot can crawl product pages, in robots.txt and your firewall.
- Prices, stock and variants are in the HTML, not loaded later by JavaScript.
- Page prices match feed prices, including sale prices and currency.
- Landing URLs work and go to the right variant, without redirects to a homepage.
Contradictions between the feed and the page are a reason for an assistant to prefer a seller whose data is consistent. Why ChatGPT may cite a homepage rather than product pages is covered in why ChatGPT cites your homepage but not your product pages.
Why are my Merchant Center products approved but missing from AI recommendations, and which competitors appear instead?
Approval in Merchant Center makes products eligible for Google Shopping, and that matters for ChatGPT too: Search Engine Land and Peec AI found 83% of products in ChatGPT's shopping carousels come from Google Shopping results, and Semrush found ChatGPT's top product was in Google Shopping's top three 75% of the time. But approval is not a ranking on either surface, so products that rank poorly in Google Shopping rarely surface in ChatGPT. AI shopping answers favour products whose data matches the question and whose quality is supported by evidence: reviews, ratings, editorial coverage and presence at trusted retailers. Profound's help documentation puts it directly: AI shopping results are not determined by organic rankings, and a competitor can win placements through feed quality, merchant relationships, structured data and product attributes.
To see who wins instead, run your buyer questions and record the products and merchants named, or use a shopping-aware tracker. Tools such as Profound and Evertune show which products and retailers appear for a category, as compared in Profound vs Evertune for retail product recommendation monitoring. Then compare the winners' listings with yours: attributes, price, reviews, and where they are sold.
Is someone else selling your product?
If large retailers or marketplaces also sell your products, ChatGPT may show their offer rather than yours. When OpenAI launched Instant Checkout in 2025, it listed availability, price, quality and whether the merchant is the primary seller or maker among the factors in which merchant appears first for the same product. If you are the maker, make that clear on your site and in the feed's brand and seller fields, keep your own price and stock competitive, and make sure your offer is available in the markets where the product appears.
What should you do, in order?
- Confirm you are onboarded to ChatGPT's feed programme, or connected through Shopify or Etsy.
- Check eligibility flags, markets and required fields, and fix invalid rows.
- Test commercial questions that trigger shopping results.
- Enrich titles, descriptions and attributes with the facts buyers ask about.
- Align product pages with the feed, and make sure OAI-SearchBot can crawl them.
- Build evidence: reviews, retailer presence and editorial coverage.
- Monitor monthly which products and merchants appear, and how yours are described.
What might this look like in practice?
Consider an illustrative cookware maker, Ironbloom, whose feed was accepted but whose skillets never appeared in ChatGPT. It turned out the team had been updating only its Google Merchant Center feed; its ChatGPT application was still on the waitlist, and ChatGPT was relying on crawled pages. Those pages loaded prices by script, and two department stores sold the same skillets with full attribute data. Ironbloom moved prices into the HTML, rewrote titles with size, material and "induction compatible", added a comparison table, and completed onboarding when invited. Within two months, its skillets began appearing for "best cast iron skillet for induction hobs".
What do people ask about product feeds and ChatGPT?
Does ChatGPT use my Google Merchant Center feed?
Not directly. ChatGPT does not read your Merchant Center account, but most products in its shopping carousels come from Google Shopping results, which your Merchant Center feed powers. OpenAI also runs its own feed programme, which accepts a Google-compatible format once it confirms it for your registered feed.
Does an accepted feed guarantee visibility?
No. It makes products eligible; whether they are recommended depends on relevance, data quality and evidence.
Why do competitors appear instead?
Usually because their data matches the question more closely, or they have stronger reviews and retailer presence.
How long after onboarding do products appear?
It varies. Test monthly with commercial questions, and check crawlability and data quality if nothing changes.
Published by The Kovmere Team · updated Sep 29, 2026 · More resources