llms.txt is a proposed standard: a markdown file at the root of a site (/llms.txt) that lists the pages most useful to a large language model, with short descriptions.
The proposal comes from llmstxt.org. It is a community proposal, not a specification from any AI provider.
What it looks like
llms-full.txt contains the full text of the listed pages rather than links.
The honest position on whether it helps
No major AI provider has publicly confirmed that it readsllms.txt for retrieval. Support is asserted more often than it is demonstrated. Treat any claim that it drives citations as unproven.
What can be said:
- It costs almost nothing. For a docs site it is usually auto-generated.
- It cannot hurt, provided it is accurate and does not expose pages you intended to keep private.
- It is not a substitute for the actual work. A well-structured, crawlable, factual page is what gets cited. A file listing badly-structured pages does not make them citable.
llms.txt as the reason for citation gains is claiming more than the evidence supports.
Does Ryze support it?
If your documentation runs on a platform that generatesllms.txt automatically — Mintlify does, for example — you already have one, and pages are included as they are published.
For a Shopify storefront, there is no automatic generation. It would need to be added as a static file at your domain root.
What to do instead, in priority order
- Make pages crawlable and server-rendered — see why your brand isn’t cited
- Make pages narrow and factual — one topic each, definition first
- Add valid structured data — see structured data for AI search
- Keep an accurate XML sitemap, which is consumed by crawlers that demonstrably matter
- Add
llms.txtif it is free to do so