Skip to main content

Making Your Documentation AI-Ready with llms.txt

· 3 min read
Mattias Sander
Mattias Sander

AI tools are changing how people find and consume documentation. But most documentation systems weren't built for AI consumption — they were built for browsers. The result: AI models struggle to extract structured knowledge from your help output, and your users get incomplete or hallucinated answers.

The llms.txt standard changes that.

What is llms.txt?​

llms.txt is a proposed standard that provides a structured index of your documentation in a format optimized for large language models. Think of it as robots.txt for AI — instead of telling crawlers what not to index, it tells AI models exactly what to read and how your content is organized.

A typical llms.txt file looks like this:

# Project Name

> Brief description of the project

## Section Name

- [Topic Title](https://example.com/docs/topic.md): Short description
- [Another Topic](https://example.com/docs/another.md): Short description

Why it matters for technical documentation​

Traditional help output — HTML with navigation chrome, JavaScript widgets, and complex layouts — is noisy for AI. Models waste context window on sidebars, footers, and UI elements instead of your actual content.

With llms.txt and companion Markdown files, you give AI models:

  1. A clean table of contents — the structure of your documentation in a machine-readable format
  2. Pure content — Markdown files stripped of layout noise, with just the information that matters
  3. Metadata — descriptions that help models understand what each topic covers before loading it

How to generate llms.txt from MadCap Flare​

The AI Helper Plugin includes a dedicated LLMS.txt Tools tab that generates these files directly from your Flare build output:

  • Generate LLMS.txt — Converts your HTML build output into clean Markdown files and creates a structured llms.txt index following your TOC hierarchy
  • Generate Markdown Target — Creates a standalone documentation package with only essential files (Markdown, images, resources), ready for AI consumption
  • Add Description — Lets you add or edit description meta tags per topic, which are automatically extracted during generation

The process is straightforward: build your Flare project as usual, then run the LLMS.txt generator. It parses your TOC structure, converts HTML to clean Markdown, and produces a properly formatted llms.txt that maps your entire documentation.

Configuration options​

The plugin gives you control over how the output is generated:

  • Auto-append .md to filenames — Choose between file.htm.md or file.md format
  • Glossary Term Handling — Control how MadCap glossary popups are processed: remove definitions, keep inline, or append as a glossary section

What this means for your users​

When your documentation has an llms.txt, AI-powered tools can:

  • Answer questions accurately by reading the actual source documentation instead of guessing
  • Reference specific topics with proper links back to your help system
  • Understand context — the hierarchy tells the AI how concepts relate to each other
  • Stay current — regenerate after each build, and AI tools always have the latest version

Getting started​

  1. Install the AI Helper Plugin in MadCap Flare
  2. Build your project normally
  3. Open the LLMS.txt Tools tab
  4. Click Generate LLMS.txt and point it to your build output
  5. Deploy the generated files alongside your documentation

The AI-readiness of your documentation goes from zero to complete in a single build step.