What is llms.txt?
llms.txt is an emerging community proposal for a Markdown file at a site's root. Its purpose is to give language-model clients a compact project summary and a curated set of useful links when context is limited. The proposal describes a required H1 project name, a following blockquote summary, optional plain Markdown context, and H2 sections containing Markdown links. A specially named ## Optional section marks lower-priority resources.
It is not a web standard, access-control mechanism, crawl directive, content license or guarantee that an AI product will read or cite the site. Client support and processing behavior vary. Publish one only when it helps a real reader or client navigate stable, public resources.
llms.txt vs robots.txt vs sitemap.xml
| File | Main job | Controls access? | Google Search role |
|---|---|---|---|
/llms.txt | Proposed, curated Markdown context for willing LLM clients | No | Google says it is not used for ranking or AI Search features |
/robots.txt | Crawler directives and Sitemap discovery | Controls compliant crawling, not authorization | Used for crawl management |
/sitemap.xml | Canonical URL discovery plus supported metadata | No | Supported discovery signal; inclusion does not guarantee indexing |
A defensible publishing workflow
- Start from canonical public pages. Export or paste a child Sitemap
urlset, then remove login, account, internal search, duplicate, redirected, noindex and low-value archive URLs. - Choose, do not dump. A useful file points to core documentation, major tools and a small number of representative guides. Hundreds of unexplained links duplicate the Sitemap and make selection harder.
- Write factual descriptions. Say what each resource contains and when to use it. Do not invent claims about rankings, freshness, security or compatibility.
- Validate and verify manually. This tool catches local structural mistakes. You must still check HTTP status, canonical tags, robots meta, page content and whether each link remains current.
- Publish at the root. Save the result as UTF-8 plain text at
https://example.com/llms.txt. Keep it in source control and review it when site structure changes.
What this local validator can and cannot prove
It can prove properties visible in the pasted text: whether the first content line is an H1, whether a blockquote summary precedes the sections, whether section bullets use parseable Markdown links, and whether exact URL strings are duplicated. It also flags HTTP URLs, fragments, common tracking parameters, private-looking paths and unusually large link lists.
It cannot prove that a URL returns 200, matches its canonical tag, is indexable, is allowed by robots.txt, has useful content or will be consumed by any AI system. Those checks require live crawling and product-specific evidence. The Sitemap importer similarly reads only pasted XML; it does not download child Sitemaps or inspect pages.
Privacy and security notes
- All builder, validator and XML parsing code runs in this browser. No content is uploaded to EasyTool.
- Do not publish private endpoints, signed URLs, tokens, internal hostnames or customer-specific links in a public root file.
llms.txtis public documentation, not a place to store instructions that replace authentication or authorization.
Primary sources
- Official llms.txt proposal and format
- Google Search: AI features and website guidance
- Google robots.txt specification
- Sitemaps XML protocol
Reviewed: 2026-08-02. Because the proposal and client support can change, verify current behavior with the consuming product before making operational claims.