How to Create an llms.txt File: Steps + Template (2026)

Part of our llms.txt guide.

To create an llms.txt file, write a Markdown file with an H1 site name, a blockquote summary, and H2 sections of curated links with one-line descriptions, then upload it to your site root at /llms.txt. It takes minutes. Set expectations accordingly: no major engine confirms using it, and studies show no measured citation lift.

Before you start: set expectations

Creating the file is easy. The honest part is what to expect from it. In July 2025 Google said it does not support llms.txt, and a study across roughly 300,000 domains found no clear effect on AI citations. Build it as low-cost hygiene, not as a visibility lever. If your CMS or docs platform can generate it automatically, even better, because then it stays current for free.

Step 1: List your most important pages

Pick the 10 to 40 pages you would want an AI model to read if it only had a small budget. For docs, that is usually the quickstart, the API reference, key concept guides, and pricing. Skip login pages, thin pages, and duplicate content. The point is curation, not a full sitemap.

Step 2: Prepare clean Markdown versions where you can

The spec encourages linking to clean .md versions of pages, so a model that follows a link gets text rather than HTML soup. If your platform can serve a page at both /quickstart and /quickstart.md, link to the Markdown. If not, link to the normal URL. This step is optional but is the main practical benefit of the whole exercise.

Step 3: Write the file

Follow the structure: one H1, one blockquote, then H2 sections. Keep descriptions to a single line and make them specific.

Step 4: A copy-paste template

[One or two sentences: what this site is, who it serves,

and what these docs cover.]

[Optional short paragraph with any context a model should

have before using the links below.]

Getting started

• Quickstart: Ship your first result in minutes

• Installation: Requirements and setup

Reference

• API reference: Every endpoint, field, and error code

• Configuration: All settings and defaults

Guides

• Common workflows: End-to-end examples

• Pricing: Plans and limits

Optional

• Changelog: Release history

• Blog: Announcements and deep dives

The ## Optional section is a real part of the spec: it flags links a model can drop first if it is short on context.

Step 5: Upload to your site root

The file must be reachable at https://yourdomain.com/llms.txt. That means the root, not a subfolder. On most hosts you drop it in the public/static directory. On Vercel or Netlify, put it in the public folder. On WordPress, use a plugin that writes root files or add it via your server config.

Step 6: Verify it is live

Open the URL in a browser and confirm you see raw Markdown, not a 404 or an HTML wrapper. Check that every link resolves. A broken link in llms.txt is worse than no link.

Step 7: Keep it current

A stale llms.txt that points to moved or deleted pages actively misleads any model that reads it. If you cannot automate generation, put a recurring reminder to review it whenever your site structure changes.

Common mistakes

Should you also create llms-full.txt?

Only if your content is text-heavy and fits in a reasonable file. llms-full.txt inlines the actual page content into one file. It suits documentation sites and is overkill for most marketing sites. See llms-full.txt for the tradeoffs.

Realistic expectations, one more time

• Time to build: minutes to an hour

• Risk: negligible

• Confirmed benefit from any major engine: none

• Measured citation lift in independent studies: none

Do it because it is cheap and tidy, not because a vendor promised results. The reliable way to earn AI citations is content that clearly answers real questions, in crawlable HTML, with structured data.

FAQ