llms.txt for ecommerce stores: what to include

Why it matters for AI agents

/llms.txt is a proposed convention for a Markdown file at the site root that tells language models and AI agents what a site is and where its important content lives. The format is simple: a required H1 with the site name, a blockquote summary, then H2 sections containing Markdown link lists.

For a store, llms.txt works as a curated entry point. Instead of crawling navigation, an AI shopping agent can learn in one request where the collections are, how search works, what the return policy says, and whether the store exposes agentic commerce endpoints (see agentic commerce protocols).

The evidence is mixed, which is why this check has the lowest weight:

• Against: Google says Search does not use AI text files like llms.txt (Google AI optimization guide), and no major assistant documents reading it.
• For: Chrome Lighthouse’s Agentic Browsing category includes an llms.txt audit, and at least one large hosted commerce platform began serving /llms.txt on its stores by default in 2026. Coding agents read it when pointed at it.

What the scanner checks

The scanner requests /llms.txt and /agents.md. A file counts as found when it returns a successful status, its content type is not text/html, and the body does not start with <!doctype or <html (this filters out storefronts that answer every path with an HTML page). If both exist, /llms.txt is graded.

Result	When
Pass	File has a # title line, at least one link (Markdown link or http(s):// URL), and mentions commerce topics: products, collections, catalog, shop, policies, shipping, returns, refunds, cart, checkout, UCP or MCP
Partial	File exists but is missing one or more of the three above
Fail	Neither file exists (or /llms.txt returns a server error)

How to fix it

Create /llms.txt with links an agent would actually use. An example for a store:

# Example Outfitters

> Outdoor apparel and socks. Ships to the US and Canada. Prices in USD.

## Catalog

- [All products](https://shop.example.com/collections/all)
- [Socks](https://shop.example.com/collections/socks)
- [Search](https://shop.example.com/search?q={query}): replace {query} with search terms

## Policies

- [Shipping](https://shop.example.com/pages/shipping): rates and delivery times
- [Returns and refunds](https://shop.example.com/pages/returns): 30-day returns
- [Contact](https://shop.example.com/pages/contact)

## For agents

- [Sitemap](https://shop.example.com/sitemap.xml)
- [UCP profile](https://shop.example.com/.well-known/ucp)

Only list endpoints you actually serve. Then:

1. Serve it at /llms.txt with HTTP 200 and Content-Type: text/markdown (or text/plain).
2. Optionally mirror it at /agents.md.
3. Keep the links current; an llms.txt that points to deleted collections is worse than none.
4. Verify: curl -i https://shop.example.com/llms.txt returns 200 and a body starting with # .

Common mistakes

• Catch-all routes that return the homepage HTML with 200 for /llms.txt. The scanner treats this as missing.
• Copying a docs-site template that links blog posts and the About page but no collections, search or policies.
• Listing endpoints that don’t exist, such as an MCP URL you never deployed.

How to fix it on Medusa

Medusa’s DTC starter storefront does not ship an llms.txt, so add one to the Next.js app:

• A static file at public/llms.txt, or
• A route handler at app/llms.txt/route.ts that returns text/markdown (Next.js route handlers). A route handler can build the collection links from the Store API, so the file stays current.

Link your regions, top categories, the search URL and your policy pages. For real-world examples of the format, see medusajs.com/llms.txt and docs.medusajs.com/llms.txt. The Medusa agent-readiness guide covers the rest of the starter’s gaps.

Sources

• llms.txt proposal
• Chrome Lighthouse: llms.txt audit
• Chrome Lighthouse: Agentic Browsing scoring
• Google Search: AI optimization guide
• Next.js route handlers