llms.txt, generated from the catalogue
The engine builds the files an AI crawler and a search crawler read: llms.txt and
llms-full.txt, a sitemap index with three sitemaps, robots.txt and a redirect table.
All of them come from the live catalogue, in each language the store serves, and the storefront serves them at
its own root.
How it works
The files are routes, not a build step. The API answers /v1/llms.txt,
/v1/llms-full.txt, /v1/sitemap_index.xml with its three children,
/v1/robots.txt and /v1/seo/redirects. The storefront proxies each one at its public path
and copies the body. Nothing is regenerated by hand when a product is published.
Language is part of the answer. The two llms files are resolved against the request's
Accept-Language and answer Vary: Accept-Language, so a shared cache keeps one copy per
language and each copy carries that language's URLs.
GET /v1/llms.txt Accept-Language: en HTTP/1.1 200 OK Cache-Control: public, max-age=3600 Vary: Accept-Language Content-Type: text/plain; charset=utf-8 # Demo Store > A demonstration store for themerchantengine.
The per-page head is the storefront's job, and the shipped one does it. JSON-LD per page type, the canonical,
the hreflang pair with x-default, and lang and dir on the page.
The documentation shows the JSON-LD one product page emits, field by field, from the product response that built
it.
Changes are announced. When a product becomes visible the engine pings IndexNow, and submits to Google Search Console when the service account is configured. The admin carries the submission history and a manual resubmission for a single URL.
Where to read the detail
- SEO and GEO: every surface, its headers, and the proxy handler that serves it.
- Change SEO and GEO output: editing what the shipped storefront emits.
- SEO API reference: the routes and their parameters.
Questions
What is llms.txt?
A plain Markdown file at the root of a site that tells an AI crawler what the site is and where its content lives, the way robots.txt tells a search crawler what it may fetch. The engine builds one from the live catalogue.
Do I have to generate it myself?
No. The API builds llms.txt and llms-full.txt from the store’s own catalogue and pages, with your storefront’s URLs inside them, and the storefront serves them at its root.
Does it follow the store’s languages?
Yes. Both files are resolved per Accept-Language and answer Vary: Accept-Language, so an Arabic request gets the Arabic catalogue and a French request gets the French one, each with the locale-prefixed URLs.
What else is generated?
A sitemap index with sitemaps for products, categories and pages, robots.txt with a crawler roster, and a redirect table the storefront honours with a 301 before it routes.
Does anything get told when a product changes?
Yes. The engine submits changed URLs to IndexNow, and to Google Search Console when you configure the service account. The admin has a screen for the submissions and their state.
Is this only for the shipped storefront?
No. The files are API routes. A storefront you build yourself proxies them at its own root, byte for byte, and the documentation shows the handler that does it.
Last updated . Every claim on this page is read from a published source and cited beside it.