# Markdown for AI agents

> Markdown for AI agents means a website can serve a page as clean Markdown, through a .md URL or content negotiation, instead of the full HTML.

- **Check ID:** A21
- **Area:** [AI Readiness](https://be1st.ai/glossary/ai-readiness)
- **Tool:** [AI Readiness](https://be1st.ai/tools/ai-readiness)
- **Page:** https://be1st.ai/glossary/ai-readiness/markdown-for-ai-agents

## Why it matters

An AI agent that downloads HTML must first strip navigation, scripts and styles, spending several times more tokens on it. The Markdown version of the same page is usually far smaller (on Cloudflare docs we measured over 90% savings), so more real content fits into the model context. This is not llms.txt: llms.txt indexes the whole site, this is the content of a single page.

## In practice

Both developers.cloudflare.com and docs.railway.com answer a request with Accept: text/markdown with clean Markdown and a Vary: Accept header, so an agent and a regular browser each get a suitable form of the content from the same address.

## How be1st checks it

Checks whether the site can serve the content of a specific page directly as Markdown. It tests content negotiation (a request carrying Accept: text/markdown), a Markdown alternative via `<link rel="alternate">` or an HTTP Link header, and the existence of a .md version of the page. The response body is validated so that HTML dressed up as Markdown does not count.

[AI Readiness](https://be1st.ai/tools/ai-readiness), Formats tab.

Every check ends as: Pass, Warning, Fail.

## Sources

- [Markdown for Agents](https://developers.cloudflare.com/workers-ai/features/markdown-conversion/) (Cloudflare)
- [HTTP Content Negotiation](https://developer.mozilla.org/en-US/docs/Web/HTTP/Guides/Content_negotiation) (MDN)

## Related terms

- [llms.txt existence](https://be1st.ai/glossary/ai-readiness/llms-txt-existence): llms.txt is a Markdown file at the root of a website that gives AI models a short overview of the site and links to its most important pages.
- [Dashes in text (AI writing tell)](https://be1st.ai/glossary/ai-readiness/dashes-in-text): Long dashes in a text (the em dash, and the en dash outside number ranges) are the best known trace of text written by a language model.
- [Organization schema](https://be1st.ai/glossary/ai-readiness/organization-schema): Organization schema is Schema.org structured data that describes the company behind a website: its name, logo, contact details and social profiles.
- [Product schema](https://be1st.ai/glossary/ai-readiness/product-schema): Product schema is Schema.org structured data that describes a product or service: its name, price, availability and reviews.
- [FAQ schema](https://be1st.ai/glossary/ai-readiness/faq-schema): FAQ schema is Schema.org structured data (FAQPage) that marks up the questions and answers on a page in a machine-readable form.
- [Review / rating schema](https://be1st.ai/glossary/ai-readiness/review-rating-schema): Review and AggregateRating schema are Schema.org structured data that describe the reviews and the average rating of a product or service.