Quantum Quotient LLC. Public scans of 95 US direct-to-consumer Shopify stores, September and October 2026.
ChatGPT, Gemini, Perplexity and Claude now recommend products and, through Shopify Catalog and Google Merchant Center, increasingly sell them. All of them read the same thing: the product data a store publishes. We scanned 95 Shopify storefronts to see how much of that data is actually there.
The short version. Not one of the 95 stores publishes complete product markup on its product pages. About half the products have no GTIN. About two thirds of product photos have no descriptive alt text. Almost nobody blocks the AI crawlers, so the assistants can read these stores; what they find is thin.
Everything a browsing assistant or a feed crawler can see without logging in: the public product listing, 30 products per store, six product pages per store for structured data, robots.txt, a live fetch as each AI crawler, the policy pages and the /.well-known/ucp profile. The checks match the public gates of the free scanner at app.youraiquotient.com/scan, and each product scores against them:
| Gate | What it means for the store |
|---|---|
| Complete product markup | Product JSON-LD on the page with rating, availability, shipping and returns. This is what ChatGPT reads when it browses a page. |
| GTIN on every variant | The barcode that lets Merchant Center and the assistants match a product to the one everyone else sells. |
| Descriptive alt text | How a crawler reads what a photo shows. |
| Substantive description | Materials, sizing, what is included. What an assistant answers from. |
| In stock | Agents skip what cannot be bought right now. |
| Image, price, brand, title | The basics a feed needs to accept the product at all. |
Scores run 0 to 100 and weight the gates by how much they decide whether a product is shown. Category and attributes are not public, so the in-app scores of the same stores would be lower.
All 95 stores fail complete product markup on most of their product pages, and 89 fail it on every page we fetched. Across the September sample, 1,208 of 1,222 products did.
Nearly every page has some Product JSON-LD, because the theme writes it. What is missing is the rest: a rating, an explicit availability, shipping details, a return policy. Among the 53 stores scanned in October, 35 showed no rating in markup on any of six product pages and 6 showed one on all six. Review apps that only draw stars on the page, and never write them into the structured data, are the usual reason.
Where the barcode field is public, 537 of 1,031 variants have none. At least 50 of the 95 stores miss it on most of their products and 21 miss it on every product we sampled. Coffee, spices, candles and pottery are the worst: house-made goods where nobody ever printed a barcode.
Merchant Center limits the visibility of products that have a GTIN and do not send one. Shopify Catalog and the assistants use the same code to work out that your product is the same item another store lists.
631 of 1,003 product images sampled in September have no alt text, or alt text under the length a crawler can use. At least 48 of 95 stores miss it on most products, 22 on all of them. Alt text is the only description of an image a text model gets.
397 of 1,222 products have a thin description. 21 stores fail on most products. The rest write real copy, which is the good news: the model that answers a shopper's question has something to answer from.
5 of 95 stores refuse at least one AI crawler. All five block PerplexityBot, ClaudeBot and Google-Extended; three also block Bingbot. The other 90 allow everything. Bot protection on large brands' CDNs is the cause in each case, not a deliberate robots.txt rule. The data problem is not access. It is what is there once the crawler arrives.
Images, prices, brand and feed-length titles pass on almost every product. Out-of-stock products are 12% of the sample. 15 of 95 stores have no shipping policy page, which Merchant Center requires. 35 of the 42 stores checked for it in September already publish a UCP profile at /.well-known/ucp, so the plumbing for agent checkout is ahead of the data that would feed it.
Stores with 1,000 or more products score a median of 69. Stores under 1,000 score 68. The gaps are the same at every size; a bigger catalogue just has more of them.
| Score | Stores |
|---|---|
| Under 50 | 3 |
| 50 to 69 | 48 |
| 70 to 79 | 28 |
| 80 and over | 16 |
Median 69, best 93, worst 41.
Our app, Product Markup for AI Shopping, does the first four from inside Shopify and the fifth every week. It drafts, you approve, and any batch can be undone. It is on the Shopify App Store. The scan behind this report is free at app.youraiquotient.com/scan.
Ninety-five storefronts, sampled from US direct-to-consumer brands in coffee and food, apparel, home, beauty, pets and outdoor, scanned 8 September and 9 October 2026. Thirty products per store (all of them for smaller catalogues), six product pages per store for structured data, one live fetch per crawler. Stores that are not on Shopify, or that refused the public product listing, were dropped. Stores are not named; the point is the pattern, not the merchant. Barcodes are only counted where the store exposes the field. The in-app check also scores category and attributes, which are not visible from outside.
Questions and corrections: support@qquotient.com.