ProductSeen AI Shopping Score: methodology
The ProductSeen AI Shopping Score (0–100) measures how easily ChatGPT and other AI shopping assistants can read a store's products. It is an independent, automated measurement. It is not a rating of product quality, service or sales, and it does not predict whether AI will recommend a store.
What we check
| Category | Points | What it measures |
|---|---|---|
| AI crawler access | 25 | robots.txt rules for OAI-SearchBot (8), ChatGPT-User (5), GPTBot (3), PerplexityBot (3), Google-Extended (2), ClaudeBot (2), Applebot-Extended (2) |
| Product data for AI | 35 | Product JSON-LD on up to 5 sampled product pages: schema present, price, availability, brand, GTIN/MPN, SKU, rating, image |
| Catalog completeness | 15 | Shopify public catalog (up to 250 products): description length, images, SKUs, product types |
| Titles, descriptions & previews | 10 | Product page title length, meta descriptions, Open Graph tags |
| Discoverability | 15 | Sitemap with products, crawlable homepage, Organization schema, llms.txt |
If a site shows our scanner a bot-challenge page, the score is capped at 35. Grades: A 90+, B 80–89, C 70–79, D 60–69, F under 60.
Reports also include checks against OpenAI's published product feed fields, a comparison with stores in the same category, and a sample rewrite. These don't change the score.
How we scan
- Our crawler identifies itself as
QuantumLabAIScan(details) and honors robots.txt. It sends about 12 requests per scan, never runs JavaScript, never logs in, and never submits forms. - Listed stores are re-scanned weekly. Every scan is kept with its scanner and method version, so scores can be compared over time.
- Categories are assigned automatically from the store's own description and product types. US state comes from the location the store publishes on Shopify.
Rankings
- Rankings are narrow (one category and/or one US state), cover US stores only, and focus on small and mid-size independent stores: sites in the Tranco top 20,000 are left out.
- Only stores scoring 70+ are listed by name. Lower scores are never published by name.
- A ranking page appears only when at least 5 stores qualify.
- Any store can be removed or corrected: see the removal and correction policy.
Changes to the method
A change to the scoring bumps the method version, and pages show which version produced each score. Version history: 2026.10-v1.3 (Oct 17, 2026), the first public ranking method.