What we audit
The same six weighted categories as the free scanner and the WordPress plugin, applied to a storefront: agents.md provenance and theme robots.txt state, product schema coverage, collection/product structure, content readiness, and authority signals. A category that genuinely doesn't apply to a storefront is marked not applicable, and the score renormalizes over the categories that remain — it doesn't penalize a store for lacking something that was never relevant.
The six categories, applied to a storefront
| Category | Status | What it checks on a storefront |
|---|---|---|
| Crawlability | Applicable | Whether your theme's robots.txt.liquid blocks known AI crawlers, and whether your agents.md is still Shopify's unmodified platform default or has been customized to actually describe your store. |
| Structure | Applicable | Heading hierarchy and content-to-noise ratio on product and collection templates — many themes bury product descriptions under dense navigation and upsell markup. |
| Entity Signals | Applicable | Product/Offer JSON-LD coverage and validity — price, availability, brand — the structured data AI shopping agents actually parse to recommend a product. |
| Citation Readiness | Applicable | Whether product and collection pages have enough original, specific detail (materials, sizing, real specs) to be quotable, versus generic manufacturer boilerplate. |
| Content | Applicable | Description depth and freshness signals across your catalog. |
| Authority | Partially applicable | Store-level trust signals (About page, contact info, policies) — some authority signals (author bylines) that make sense for editorial content don't apply to a storefront the same way. |
When a category doesn't apply
Not every check that makes sense for a blog post or documentation page makes sense for a product listing. When a category is judged not applicable to a given page type, it's excluded from that page's score entirely — the remaining categories' weights are renormalized so they still sum to 100%, the same exclude-rather-than-penalize approach the scanner uses for checks it can't determine.
agents.md provenance, specifically
Because every Shopify store gets an identical default agents.md (see the overviewfor why no app can change that), a bare "present" check would be true but useless — it would pass every uncustomized store identically. The audit instead checks whether your agents.mdstill matches Shopify's known default boilerplate (the same detection the free scanner already runs) and treats an unmodified default as a finding, not a pass, since it describes the platform, not your store.
Related
- Install & permissions
- The AI Readiness Engine — the same six-category model, documented in full for the npm package.