Markdown mirror - every page available as plain Markdown by adding .md
Tests whether your HTML pages also resolve at the same URL with a .md suffix, returning clean Markdown.
What this signal tests
We check whether your pages are mirrored as Markdown at the same URL plus a `.md` suffix - for example, /docs/getting-started also resolves at /docs/getting-started.md, returning the same content as plain Markdown instead of HTML. It is a proposed convention from the llms.txt ecosystem, intended to let AI fetchers skip HTML parsing entirely.
Why it matters for your visibility in AI
Every AI crawler today has to download HTML, strip ads and navigation, decode JavaScript, and try to recover the underlying article. It is error-prone - models routinely hallucinate "Buy Now" buttons into article text, lose code blocks, or mangle tables. A Markdown mirror skips all of that: the model gets exactly the content you wrote, in the structure you wrote it, with no chrome to misinterpret. For a documentation site, a knowledge base, or any text-heavy publication, this means measurably more accurate AI answers about your content. The consequence is concrete: when an AI cites a passage from your site, the citation is far more likely to be exact rather than paraphrased with subtle errors. The convention is still niche, but the upside-to-effort ratio is high for any site already authored in Markdown.
Pass criteria at a glance
| Criterion | Passes when |
|---|---|
| >=80% sampled URLs return Markdown. |
How we test it
We take a sample of URLs from your sitemap (or a crawl) and for each one, request the same URL with `.md` appended. We expect a 200 response with Content-Type text/markdown (or text/plain) and a body that is recognisably Markdown - heading syntax, link syntax, no opening `<html>` tag. We pass the signal if at least 80% of sampled URLs return valid Markdown. Anything lower and we treat it as not implemented, since partial coverage confuses crawlers.
Show technical detection method
For a sample of crawled URLs, append .md and GET; pass if >=80% return 200 with text/markdown.
If your site fails: how to fix it
- If you use Mintlify, Docusaurus, or Nextra, enable the built-in Markdown mirror - most modern docs platforms ship this as a one-line config flag.
- If you have a custom build pipeline, add a route that takes any `/path.md` request, looks up the same path in your Markdown source tree, and returns it with Content-Type: text/markdown; charset=utf-8.
- Strip frontmatter (YAML at the top of source files) before serving, so the AI sees content, not build metadata.
- Ensure your sitemap still lists the canonical `.html` URLs - the .md versions are a parallel address, not a replacement.
- Test with a sample fetch. The convention is described at https://llmstxt.org/. Because it is a proposed convention, have a developer who can read the source format implement it.
Quick facts
| Maturity | PROPOSED |
|---|---|
| Weight | low |
| Category | Emerging Standards |
Primary sources
Related signals
Frequently asked questions
Is this a real standard or just a proposal?
It is a proposed community convention from the llms.txt ecosystem, not an IETF or W3C standard. Adoption is concentrated among docs platforms (Mintlify, Nextra, some Docusaurus sites). It is one of the less-mature signals in this category.
Which AI systems use the .md suffix today?
Mostly developer-facing tools: Cursor, Continue, Claude Code, and some Perplexity flows. Consumer chat assistants rarely check for it yet. It is a forward-looking optimisation more than a present-day necessity.
Will fixing this still matter in two years?
Plausibly yes, but less critical than llms.txt or MCP. If AI fetchers get dramatically better at HTML-to-Markdown conversion, the value of a dedicated mirror drops. For now it is a low-cost hedge if your site is already Markdown-authored.
Is there an easier alternative?
Yes - publishing a good /llms-full.txt covers most of the same ground and is one file instead of a routing change. If you can only do one, do llms-full.txt first.
Run your own scan
Run a free scan and see how your site grades across all 155 AI-readiness signals.