Markdown versions of your pages for AI agents: why and how
Updated 10 Nov 2026 · by Venu Gopal Reddy Muvva, founder of BrandInAIAnswers · official sources checked 10 Nov 2026
A Markdown twin is a clean text version of a page served at the same URL with .md added (for example /pricing.md), so AI agents can read the content without menus, scripts or layout. The llms.txt proposal recommends this convention and suggests linking to twins from llms.txt and with rel="alternate" type="text/markdown".
The convention
- Same URL plus .md: /pricing → /pricing.md; for URLs without a file name, index.md.
- Discoverable with <link rel="alternate" type="text/markdown" href="/pricing.md"> in the page head, or an HTTP Link header.
- Listed in llms.txt so agents find the important ones first.
- Optional: serve Markdown when a request sends Accept: text/markdown.
Example
<!-- in /pricing -->
<link rel="alternate" type="text/markdown" href="https://example.com/pricing.md">
# /pricing.md
---
title: Pricing
canonical: https://example.com/pricing
---
# Pricing
| Plan | Price | Users |
|---|---|---|
| Starter | ₹499/mo | 1 |
| Team | ₹1,499/mo | 5 |Put a canonical link back to the HTML page so search engines treat the HTML as the main version.
Ways to produce twins
| Approach | Good for |
|---|---|
| Write content in Markdown and render both | Docs sites, blogs |
| Convert rendered HTML to Markdown on request | Existing sites; keeps twins identical to the page |
| Export at build time | Static sites |
What a good twin contains
The same headings, paragraphs, lists and tables as the page, in the same order; links as absolute URLs; the page title and a short summary at the top; and the date the content last changed.
What it leaves out: navigation menus, footers, cookie banners, scripts, tracking pixels and buttons that do nothing outside the browser. A twin of a pricing page should read as a document on its own, so an agent that reads only the twin can answer 'how much does it cost and what is included'.
Generate twins from the same source as the page, so they never drift apart; a twin that disagrees with the HTML is worse than no twin.
How to check your twins
Open a twin in a browser and read it as if you had never seen the site: can you answer the page's main question from the twin alone? Check that tables survived as Markdown tables, that links are absolute, and that the response has a text/markdown content type and a canonical Link header pointing at the HTML page.
Then fetch a few twins listed in your llms.txt with curl to confirm they return HTTP 200, not a redirect to the HTML page.
Questions
Will Markdown twins create duplicate content problems?
Point a canonical (link tag or HTTP Link header) at the HTML page and keep twins out of your sitemap.
Which pages need twins?
The pages you list in llms.txt: pricing, product, docs, policies. Not every page.
Does this site do it?
Yes: add .md to any page URL here, and see our llms.txt.
See which AI answers name you today
BrandInAIAnswers's free audit checks whether AI crawlers can read your site and asks an AI with web search the questions your buyers ask, then shows who gets named. The first 14 days of the full app are free.
Sources
- llmstxt.org: the llms.txt proposal (checked 10 Nov 2026)