What is llms.txt?

llms.txt is a text file placed at the root of your website that provides AI engines with a curated summary of your site's most important content. Think of it as a hybrid of robots.txt and sitemap.xml, optimized for large language models. It is a proposed standard (not yet universally adopted by AI engines) that aims to give AI tools a faster, cleaner path to understanding a site than crawling raw HTML. In 2026 it is "worth adding, not worth obsessing over".

The format

The file lives at yourdomain.com/llms.txt and is plain markdown. The recommended structure (llmstxt.org):

# Site Name

> One-line description of what the site is about.

## Docs

- [Doc title](https://yourdomain.com/doc-slug): short description
- [Another doc](https://yourdomain.com/another): short description

## Optional

- [Secondary links](https://yourdomain.com/secondary): short description

The file is human-readable markdown, so you can write it by hand in 20 minutes for a small site.

What llms.txt does

Three things it tells an AI engine:

  1. What the site is. The one-line description and title. Sets context before the AI crawls any pages
  2. What pages matter most. The curated link list. Not every URL on your site, just the ones you want prioritized
  3. Short descriptions per page. Helps the AI know what each linked page contains without having to fetch it

In practice: the AI engine reads llms.txt and gets a fast overview before deciding which pages to actually crawl.

What llms.txt does not do

Four things it is NOT.

  1. It is not enforced. AI engines may or may not use the file. OpenAI, Anthropic, and Perplexity have not officially committed to reading it, though some tools and third-party agents do
  2. It is not a ranking signal directly. You will not see a citation boost purely from adding llms.txt today. The benefit, if any, is making your site easier for AI to understand
  3. It is not a replacement for schema markup or sitemap.xml. Schema is machine-readable by AI engines already. Sitemap.xml is widely respected. llms.txt is additive
  4. It is not a paid-inclusion backdoor. You cannot bid for priority in llms.txt. It is a static file you write once

Is it worth the effort?

Short answer: yes, for the time investment. A 20-minute setup with a small ongoing maintenance cost.

Longer answer: the adoption curve is uncertain. In 2026, some smaller AI tools and agents parse llms.txt. The big three engines (ChatGPT, Claude, Perplexity) have not confirmed they read it as a first-class signal. Google has not indicated they will (searchengineland.com).

BUT: the file costs ~20 minutes to write and is future-proof. If adoption grows, you are already set up. If it does not, you lost 20 minutes.

How to write one for a small business

A 5-step process.

  1. Pick your top 20 pages. The pillar pages, the best-performing spokes, the key service pages, the about page
  2. Write a 1-sentence description for each. Plain English. What does this page answer
  3. Write a 2-sentence site description at the top. What is the business, who is it for
  4. Format as markdown following the schema from the spec
  5. Upload to yourdomain.com/llms.txt. Most CMS platforms allow uploading static files to the root. Static site generators make this trivial

Full setup time: 20 to 45 minutes for a small site.

Example for a service business

# Acme Marketing Consulting

> Independent marketing consultancy for B2B SaaS companies seeking organic growth.

## Docs

- [What is AI SEO?](https://acmeconsulting.com/ai-seo/what-is-ai-seo/): plain-English explainer of AI SEO for non-technical business owners
- [Our Services](https://acmeconsulting.com/services/): scope, timeline, and deliverables for our core engagements
- [AI SEO for SaaS case study](https://acmeconsulting.com/case-studies/saas-ai-seo/): 12-month growth story with data

## Optional

- [About the owner](https://acmeconsulting.com/about/): background and working style
- [Contact](https://acmeconsulting.com/contact/): book a call or send a message

That is the whole file. The business curates what matters and the AI engines get a fast path to it.

The robots.txt relationship

llms.txt complements robots.txt, does not replace it.

  • robots.txt tells crawlers what NOT to index
  • sitemap.xml lists every URL you want indexed
  • llms.txt tells AI engines which pages matter MOST

If you have all three, they work together. If you only have time for two, prioritize sitemap.xml and schema markup. llms.txt is the nice-to-have, not the base layer.

What not to do

Three mistakes.

  1. Listing every URL on your site. Defeats the point. llms.txt is a curated list, not a dump
  2. Writing marketing copy as descriptions. "The best marketing consultancy in the world" is not what AI engines want. Plain factual summaries work better
  3. Forgetting to update it. When you add important new pages or retire old ones, update llms.txt. Quarterly review is plenty