← All checks

Markdown content negotiation for AI agents

We request your homepage with the header Accept: text/markdown, and separately request /index.md (or the page path plus .md).

Why it matters

Markdown carries the same content as HTML with a fraction of the tokens, and agents that ask for it get clean text without navigation markup. Without it the agent has to parse HTML itself. A markdown response to the Accept header scores 4; only a separate .md file scores 2.

How to fix it

1. Return Content-Type: text/markdown when the request has Accept: text/markdown.
2. Add Vary: Accept so caches keep the HTML and Markdown variants apart.
3. Also publish a .md twin of each page, for example /index.md for the homepage.
4. Generate the Markdown from the same source as the HTML so they never differ.

How to verify

Run curl -s -H "Accept: text/markdown" -D - https://example.com/ | head and look for content-type: text/markdown.

Check your site free

Scan a site

Related checks

• llms.txt
• llms-full.txt / .md of key pages
• No walls
• Content without JS
• Headings

All checks