Does llms.txt work, and do I need one?

What is llms.txt?

llms.txt is a plain-text file, written in Markdown, that you place at the root of your website so it loads at yourdomain.com/llms.txt. Jeremy Howard of Answer.AI proposed it in 2024 as a short reading list for AI tools: who you are and which pages matter most. It is a proposal that anyone can follow, not an official web standard.

It is easy to confuse with two older files. robots.txt tells crawlers what they may fetch. sitemap.xml lists your pages for search engines. llms.txt does neither job: it is a curated summary, and it cannot allow or block anything.

Does llms.txt work?

• Google. Statements reported in mid-June 2026 have Google saying it does not use llms.txt for Search, with John Mueller comparing it to the keywords meta tag. Its May 2026 guide to AI features adds that no special files, markup or Markdown copies are needed for AI Overviews or AI Mode.
• Real-world use. Ahrefs studied 137,210 domains and found that 97% of llms.txt files received zero requests in May 2026. Of the requests that did arrive, under a fifth came from named AI tools.
• Answer engines. As of the first quarter of 2026, no major AI answer engine had publicly confirmed that it reads llms.txt to decide what to cite.

So the honest answer is no: there is no evidence that adding one raises rankings or AI citations. That is an absence of proof and not proof of harm, and some developer tools and agents do fetch the file, as the section on reasons to publish one explains.

How do I write one?

The format is simple: a title, a one or two sentence summary, then links grouped under headings, each with a short note. For a dental clinic the file could contain these lines, one per line:

• # Acme Dental
• > Family dental clinic in Ahmedabad offering check-ups, braces and implants.
• ## Key pages
• - [Treatments](https://example.com/treatments): what we offer, with typical prices
• - [Book a visit](https://example.com/book): online booking and clinic hours
• ## Optional
• - [Privacy policy](https://example.com/privacy)

1. 1
   Use facts already on your site

   Write the summary and notes from what your pages already say. Link only public, live pages.

2. 2
   Save it as plain text

   Name the file llms.txt and publish it at the root of your domain. Open it in a browser to confirm it loads without a login.

3. 3
   Keep it current

   List the pages you would hand a new customer or developer, and update the file when those pages change.

A generator drafts the file from your sitemap, and a checker confirms it loads and its links work. Read whatever a generator produces before you publish it. Some documentation platforms also publish a longer llms-full.txt; that is optional too.

Why publish one anyway?

• Developer and agent tooling. Coding assistants and AI agents pointed at a site, especially documentation, can read a tidy summary instead of wading through navigation. This is where the file has real traction.
• Documentation sites. If you publish product or API docs, it helps people who feed your docs to their AI tools.
• Near-zero cost. It takes about ten minutes and there is no known penalty.
• A useful audit. Choosing your key pages often exposes pages that are missing, outdated or contradictory, and fixing them helps people and search engines too.

These are housekeeping benefits, not ranking reasons. If a major engine adopts the file later, you will already have one.

What matters more than llms.txt?

Google says the same SEO fundamentals apply to its AI features: pages that are crawlable, indexable, helpful and clear. Spend your time there first. How AI assistants decide which sites to cite and how to appear in Google AI Overviews cover what has evidence behind it, and your robots.txt rules for AI crawlers decide whether assistants can reach you at all.

The Arcline website publishes an llms.txt for the modest reasons above, not as a ranking tactic. The product also checks on every scan whether your site has one and, on websites connected through GitHub, can write one from your own pages as a pull request you review. It says plainly that not every AI system reads the file. See the features page.