The web, spoken fluently to AI agents.
content-md is an open specification for high-fidelity textual representation for AI agents.
Built by a team that create documents every day
title: Introducing Content-md
description: >-
AI agents should be first-class visitors,
let's give them a tailored experience.
date: 2026-04-29
author: Alessio
license: CC-BY-4.0
---
# Introducing Content-md
AI Agents are increasingly browsing the web on behalf
of humans. The web was built with humans in mind that
demand quality and pleasant interaction. Agents go
straight to the point and prefer a more structured approach.
## The Problem
Converting complex HTML pages with navigation, ads,
and JavaScript into LLM-friendly plain text is both
difficult and imprecise.
---
description: >-
AI agents must be considered as first-class visitors,
let's give them a tailored experience.
title: Introducing Content-md
date: 2024-01-15
author: Jane Smith
license: CC-BY-4.0
---
# Introducing Content-md
AI Agents are increasingly browsing the web on behalf of humans...Frontmatter
Serves as an introductory summary — ~100 tokens, ~540 characters. AI agents read this first to decide if the full document is relevant before fetching it. Functions as a lightweight preflighted index.
Markdown body
CommonMark or GitHub-flavored Markdown. Must open with a first-level heading. Prefer text over images — link images and include alternate text. Preserve document hierarchy starting from level two.
Why content-md
AI agents can read HTML or complex formats. Time and tokens set the rules.
Use tokens wisely
The creator wins
Tools & plugins.
These tools help you start serving content-md without building from scratch.
Caddy Content Negotiation
WordPress Post to Markdown
vs. LLMs.txt, Agents.md and Skills.
Content-md vs. LLMs.txt
Covers the whole website: one URL listing everything available. Think of it as a sitemap.
→ Predictable URL at website root
→ Birds-eye view of all content
They coexist like sitemaps and pages. content-md describes individual resources.
Content-md vs. Agents.md
Targets coding agents with README context for code repositories: build steps, tests, conventions.
→ Instruct coding agents
→ Repository-scoped
content-md does not target coding agents. The two serve entirely different contexts.
Content-md vs. Skills
Provides additional knowledge and a birds-eye view of available content to agents, packaged as folders.
→ Not discovered via direct URL
content-md responses are nearly compatible with Skills; the frontmatter fields map closely.
The web, spoken fluently to AI agents