Guide
llms.txt example
The file itself, which you can open on this site; the format rules in the proposal's own words; what Google says about whether you need one; and the ten-minute version if you decide to write one anyway. Nothing on this page claims the file moves a ranking, because nobody who runs a search engine or an assistant says it does.
The work that does get a page cited is on the ChatGPT SEO guide.
The example, shortened, and the real one
This is the shape of the file, cut to a few lines so the parts are visible. The full version for this site is at passieon.com/llms.txt, generated from the same route map as the sitemap, every line the page's own title and description, so it cannot say something the site does not.
# Passieon > SEO for businesses that want customers from Google, from the AI answers people ask first, and from the map. Clinics, trades, professional services, startups and agencies, anywhere Google is the search engine people use. ## Services - [what the SEO work is](https://passieon.com/seo-services/): The work, month by month, and what it is measured by. - [local SEO, the map pack and the profile](https://passieon.com/local-seo-services/): The Business Profile, reviews and the pages for the places you serve. - [technical SEO, crawl, index and speed](https://passieon.com/technical-seo-services/): Crawling, indexing, speed and structured data, fixed and verified. ## Guides - [the SEO audit checklist](https://passieon.com/seo-audit-checklist/): Indexing, crawling, on-page, content, speed and links, in the order that finds the expensive faults first. - [how ChatGPT decides what to cite](https://passieon.com/chatgpt-seo/): From OpenAI's and Google's own documentation. ## Optional - [the blog](https://passieon.com/blog/): Articles and templates.
The format, from the proposal itself
The proposal was published by Jeremy Howard on 3 September 2024, with a second version on 10 August 2026, at llmstxt.org, and it is short. The file is Markdown rather than XML or JSON, which the proposal calls unusual and chooses on purpose, because the readers are language models and people. It lives at the root of the domain, or at a deeper path covering the pages under it.
- An H1 with the name of the project or site. In the proposal's words, this is the only required section.
- A blockquote, one short summary of the site, with the facts a reader needs before the rest.
- Any number of ordinary paragraphs or lists with more detail.
- Sections under H2 headings, each a list of links: a required Markdown link, then optionally a colon and a note.
- A section headed Optional, by convention the links an agent can skip when a shorter context is needed.
That is the whole specification. The notes after each colon are the part worth writing well, because they are what a reader with no time actually gets.
Whether you need one, from the vendors
Google's documentation on its AI features says you do not need to create new machine-readable files, AI text files or markup to appear in AI Overviews or AI Mode. It lists what it does need, and it is the ordinary list: a page that can be crawled, that is indexed, that is eligible to be shown with a snippet, and that is findable through internal links. So for Google the file is neither required nor mentioned as a signal.
The proposal's own page says thousands of sites publish one, that documentation platforms generate one automatically, and that OpenAI, Anthropic and Gemini publish llms.txt files for their own documentation. Publishing a file for your documentation is a different thing from reading everyone else's, and no vendor documentation this page could find says an assistant fetches llms.txt when it answers a question. If that changes, this page will say so with the source named.
So the case for the file is that it costs ten minutes and cannot hurt, and the case against spending an afternoon on it is everything above. Ours exists for the reason this page exists: as a live example, generated, at no cost. The work that does get a page quoted is the plain answer under a heading that names the question, which is the whole of the on-page SEO checklist and section one of the ChatGPT guide.
The ten-minute version
- Open a plain text file called llms.txt. Put the business name on the first line after a hash and a space.
- On the next line, after a greater-than sign, write the sentence you would say to a stranger about what the business does and where.
- Under a heading for services, list the service pages, each as a Markdown link followed by a colon and the page's own description.
- Under a heading for guides, list the pages that answer the questions customers ask before they buy.
- Under a heading called Optional, list everything else worth knowing about, such as the blog.
- Upload it to the root of the site and open your-site.com/llms.txt in a browser. It should show as plain text, not download.
- Keep it in step with the site. If a page changes its title, change the line; a file that names pages that no longer exist is worse than none. Ours is generated from the route map at every build for that reason.
Beside this page
The file that does decide what an AI crawler may read is robots.txt, one line per user agent, which the robots.txt generator writes with each crawler named. What the crawlers behind ChatGPT and Google's AI features actually need, from the vendors' own documentation, is the ChatGPT SEO guide.
Questions about llms.txt
What is an llms.txt file?
A plain text file in Markdown at the root of a website, proposed by Jeremy Howard in September 2024, that names the site in a heading, summarises it in a quoted line, and lists the pages an AI system should read first, each as a link with a note. The proposal's own page says the only required part is the heading; everything else is optional.
Does Google use llms.txt?
Google's documentation on its AI features says you do not need to create new machine-readable files, AI text files or markup to appear in AI Overviews or AI Mode, and lists the ordinary requirements instead: crawlable, indexable, eligible for a snippet, with the content findable through internal links. That is the vendor's own answer, and this page does not improve on it.
Which AI companies support it?
The proposal's page says thousands of sites publish one, that documentation platforms generate one automatically, and that OpenAI, Anthropic and Gemini publish llms.txt files for their own documentation. Publishing a file is not the same as reading everyone else's, and no vendor documentation this page could find says its assistant fetches llms.txt when answering. That is the honest state of it as of the date on this page.
Where does the file go?
At the root of the domain, so it answers at your-site.com/llms.txt, served as plain text. The proposal also allows one at a deeper path covering the pages under it. It is not referenced from robots.txt or the sitemap; an agent that wants it asks for the address directly.
What should it list?
The pages you would hand a new employee on day one: what the business does, the services, the guides that answer the questions customers ask, the contact page. Each line is a Markdown link followed by a colon and a one-line note, and the note is the part worth writing, because it is what a reader with no time gets. Ours lists every indexable page with its own title and description, generated from the same route map as the sitemap, so it cannot drift from the site.
Is it worth doing?
It costs ten minutes and cannot hurt, and that is the whole case for it. The case against spending more than ten minutes is that no search engine or assistant documents reading it. Write the file if you have the ten minutes, and spend the rest of the afternoon on the answer paragraph under each heading, which is what the assistants do quote.
Book a call with us
A 30-minute call in your working hours. Tell us what you sell and where, and we will tell you what we would do first and what it involves. Prefer email? Send your website through the form and the Passieon Team will reply within two working days.