How to Diagnose Product Feed Gaps in ChatGPT Shopping With Manual Searches

Where does ChatGPT get product data?

Four routes matter:

• Your own product pages. OpenAI says a feed is not required if ChatGPT can crawl your site, and Profound's July 2026 study of about 812,000 ChatGPT product cards found 87.3% came from web crawl and 12.7% from merchant feeds. Check that OAI-SearchBot is allowed in robots.txt and your firewall, and that product pages show price, stock and key attributes in the HTML.
• OpenAI's merchant programme. Merchants can apply to share product feeds with ChatGPT. OpenAI's feed specification lists nine required fields: item ID, title, description, URL, brand, seller name, image URL, price and availability. It also accepts Google-compatible feeds.
• Shopify's agentic storefronts, which make eligible Shopify stores' products available to ChatGPT through Shopify Catalog, on by default.
• Google Shopping data. Research by Semrush found that ChatGPT's product carousels draw on shopping searches. In 100 runs, the top ChatGPT product appeared in Google Shopping's top three 75% of the time. Search Engine Land reported that 83% of carousel products came from Google Shopping data. So your Merchant Center feed matters for ChatGPT too.

How ChatGPT orders shopping results is covered in how ChatGPT ranks shopping results.

How do you run the manual test?

1. Write 20 to 40 buyer questions across five types: category ("best cordless vacuum for pet hair"), attribute ("waterproof hiking boots wide fit under $150"), brand ("[your brand] mattress review"), comparison ("[your product] vs [competitor]") and availability ("[product] in stock", "pick up today near [city]").
2. Use a clean session: a temporary chat, signed out where possible, with search on.
3. Run each question three times, because results vary.
4. Record for each answer: whether a product carousel appears, whether your product is in it and in what position, the price, stock and variant shown, the merchant and link destination, and competing products.
5. Capture the search queries ChatGPT runs. Browser developer tools can reveal the fan-out queries behind an answer, as described in how to use browser tools to reveal ChatGPT's search queries and fan-outs.
6. Repeat monthly, and after any feed change.

What does each symptom mean?

OpenAI's specification notes that an omitted shipping price is read as unknown, not free. Return details apply only when you state that returns are accepted.

How to test ChatGPT product recommendations for a Shopify store

1. Confirm agentic storefronts are active in your Shopify admin, and that your products are eligible.
2. Check the Google & YouTube app for Merchant Center errors, since Google Shopping data feeds ChatGPT's carousels.
3. Run your question set as above, focusing on your best sellers.
4. Check orders for ChatGPT attribution in the Shopify admin, and chatgpt.com in sales by referrer.
5. Fix product data in Shopify, using titles, descriptions, metafields and variants, since it flows to every channel.

Independent home decor retailers checking AI visibility for high-intent buyer questions

Home decor buyers ask specific, constraint-heavy questions, and those are where feed gaps show. Test questions like:

• "Washable 8x10 rug that's pet-friendly under $300"
• "Small-space sofa bed that fits through a narrow doorway"
• "Oak dining table that seats six, delivered within two weeks"

If you sell a matching product but it does not appear, the usual culprit is missing attributes. Dimensions, materials, care instructions, delivery times and whether an item is washable or pet-friendly often live only in images or long descriptions. Put them in titles, structured attributes and clear product copy. Independent retailers also compete with marketplaces, so accurate brand, seller and GTIN data helps assistants credit you rather than a reseller.

How can an omnichannel director see whether ChatGPT recommends our ecommerce brand for click-and-collect questions?

Test the questions shoppers ask when they want an item today: "where can I buy [product] and pick it up today in [city]?", or "[product] click and collect near me". Record which retailers ChatGPT names, whether it mentions pickup, and which sources it cites.

OpenAI's discovery feed specification centres on online product data. Its field groups cover product basics, identity and variants, prices, shipping, returns, reviews, markets, ads and checkout. Pickup details should therefore also be clear on the sources assistants read. Google Merchant Center supports local inventory with pickup method and pickup time attributes, which feed Google's shopping surfaces. Your store pages and Business Profiles should state pickup options, stock by store where possible, and opening hours. If ChatGPT recommends competitors for pickup questions, check whether their store pages state pickup and stock more clearly than yours.

How to test a product feed against ChatGPT integration specs and UCP requirements

For OpenAI's feed:

1. Validate the nine required fields on every row: item ID, title, description, URL, brand, seller name, image URL, price and availability.
2. Check value formats: prices with currency, such as "18.00 USD"; availability as one of in_stock, out_of_stock, pre_order, backorder or unknown; GTINs of exactly 8, 12, 13 or 14 digits with a valid check digit.
3. Keep IDs stable when price, stock, title or images change, and never put price in an offer ID.
4. Make URLs and images public, since OpenAI requires publicly accessible product and image URLs.
5. Check flags: search eligibility is on by default, checkout is off unless separately enabled, and setting search eligibility to false also removes checkout eligibility.

For Google's Universal Commerce Protocol (UCP):

1. Publish your business profile at /.well-known/ucp, publicly accessible without authentication, as Google requires.
2. Declare only the capabilities you support, such as checkout types, fulfilment and discounts.
3. Keep Merchant Center data consistent with the profile and your site.

Use a development store or a test feed first, then compare what ChatGPT and Google's AI Mode show for your test questions before and after.

What might a diagnosis look like?

Consider an illustrative outdoor gear retailer whose waterproof jackets never appeared for "waterproof jacket for hiking under $200". Its manual test showed a reseller's listing of its own jacket instead, at an old price. The retailer's feed lacked GTINs, and listed the seller as the parent company's legal name, which did not match the brand. Its Merchant Center feed updated only daily, while prices changed during sales. The retailer added real GTINs, aligned brand and seller names, and moved to more frequent feed updates. It also added "waterproof", "breathable" and weight to titles. Within a month, its own listing appeared in two of three test runs, at the current price.

What do people ask about ChatGPT Shopping feed gaps?

Do I need to join OpenAI's merchant programme?

Not necessarily. Shopify stores reach ChatGPT through agentic storefronts, Google Shopping data also feeds its carousels, and ChatGPT can read your product pages directly if OAI-SearchBot is allowed. Joining gives more direct control.

How often should feeds update?

As often as prices and stock change. Stale data is the most common cause of wrong answers.

Can I see my feed's status inside ChatGPT?

No. Manual tests, merchant programme tools and Merchant Center diagnostics are how you check.

Why does a reseller appear instead of my store?

Usually because its listing has clearer identifiers or better data. Submit accurate brand, seller and GTIN data.