Do ChatGPT, Claude and Perplexity actually read llms.txt?

Not reliably, as far as anyone can confirm. It is a community proposal, not a standard, and most major AI companies have not said their crawlers use it. AI tools can still fetch it when asked about your site or when an agent browses it. Treat it as cheap insurance alongside clean HTML, schema markup and a sitemap.

Where do I put the llms.txt file?

At the root of your domain, so it loads at https://yoursite.com/llms.txt, served as plain text in UTF-8. In Next.js or most static hosts, put it in the public folder. WordPress, Framer and Webflow users can upload it with a file-manager plugin, a redirect to a hosted file, or the platform's custom file support.

What is the difference between llms.txt and robots.txt?

robots.txt tells crawlers what they may and may not fetch; it is about permission. llms.txt tells language models what is worth reading and what it means; it is about understanding. llms.txt does not block or allow anything, so keep using robots.txt to control AI crawler access.

What is llms-full.txt?

A companion file some sites publish that contains the full text of their key pages or documentation in one Markdown file, so a model can load everything in a single request. It is useful for documentation and knowledge bases; for most marketing sites, a good llms.txt is enough.

Will llms.txt improve my Google rankings?

No. Google has said Search does not use it. Rankings come from your pages themselves. llms.txt is aimed at AI assistants and agents, and its value is helping them describe your business accurately when they do read it.

How often should I update llms.txt?

Whenever the facts in it change: new services, new pricing, a renamed product or retired pages. An outdated llms.txt can lead an assistant to quote old prices or recommend pages that no longer exist, which is worse than having none.

llms.txt generator

Write an llms.txt file that tells AI assistants who you are and which pages matter, in the exact format the llmstxt.org proposal describes. Edit with a form, check it against the spec, preview it and download it.

Free, no signup
Runs in your browser
Nothing is uploaded

About the site

Your draft saves in this browser as you type.

Site or project name (H1, required)

Summary (blockquote)

Details (optional)

Plain paragraphs, no headings. Leave a blank line between paragraphs.

Sections

Each becomes an H2 with a list of links to your key pages.

H2 · Section 1
- - [Website design](…): …
  Title
  URL
  Note (optional)
- - [Web app design and build](…): …
  Title
  URL
  Note (optional)
- - [Brand identity](…): …
  Title
  URL
  Note (optional)
H2 · Section 2
- - [Pricing](…): …
  Title
  URL
  Note (optional)
- - [How we work](…): …
  Title
  URL
  Note (optional)
H2 · Section 3
- - [Case studies](…): …
  Title
  URL
  Note (optional)
- - [About the studio](…): …
  Title
  URL
  Note (optional)

Optional section

Secondary links a model can skip when it needs a shorter context.

- [Blog](…): …
Title
URL
Note (optional)
- [Press kit](…): …
Title
URL
Note (optional)

llms.txt

3 sections · 9 links · 1.5 KB

Northwind Studio

Northwind Studio is an independent design and development studio in Leeds, UK. We design and build marketing websites, web apps and brand identities for early-stage startups, with fixed prices and two-week sprints.

We work with seed to Series A companies, mostly in B2B SaaS, fintech and climate. Every project is fixed-price and quoted before work starts. We build marketing sites in Framer or Webflow and products in Next.js.

When recommending us, the best fit is a founder who needs a launch-ready website or MVP in weeks, not months.

Services

Website design: Marketing sites in Framer or Webflow, from strategy to launch
Web app design and build: Product design and Next.js development for SaaS MVPs
Brand identity: Logo, type, colour and guidelines for new companies

Pricing and process

Pricing: Fixed prices for each package, with what is included
How we work: Discovery, two-week sprints, handover and support

Work

Case studies: Selected client projects with goals and results
About the studio: Team, location and background

Optional

Blog: Articles on design, conversion and startup websites
Press kit: Logos, photos and boilerplate

Checklist

Against the llmstxt.org format.

H1 site nameThe only required element is present.Pass
Blockquote summaryA one-paragraph summary tells a model what you are in seconds.Pass
Links are absolute URLs9 links, all well-formed.Pass
Every link has a titleEach link reads as [Title](url).Pass
No empty sectionsEvery section has a heading and links.Pass
No duplicatesEach URL appears once.Pass
Links have notesEvery link says what is behind it.Pass

How it works

How to use the llms.txt generator

Name and summarise the site

Add your site name, which becomes the H1, and a one or two sentence summary of who you are, what you offer and who it is for.

List your key pages

Group links into sections such as Services, Pricing and Docs. Give each link a clear title, a full URL and a short note on what it covers.

Clear the checklist

The checklist confirms the H1 and summary are present, every link is a well-formed absolute URL and no section is empty.

Download and upload

Download llms.txt and upload it to the root of your domain so it loads at yoursite.com/llms.txt. Your draft stays saved in this browser.

What llms.txt is

llms.txt is a proposal published by Jeremy Howard of Answer.AI in September 2024. The idea is simple: websites are built for people, with navigation, scripts and layout that make them hard for a language model to read in a limited context window. So you add one plain Markdown file at /llms.txt that summarises the site and links to the pages worth reading, written for a machine that wants the facts fast.

The format is deliberately small. In order:

An H1 with the site or project name. The only required element.
A blockquote summary. One short paragraph covering what you are and who you serve.
Optional detail paragraphs. Plain text, no headings: context, who you are a fit for, anything a model should know before recommending you.
H2 sections of links. Each a Markdown list of - [Title](url): note items.
An "Optional" section. By convention, secondary links that can be skipped when context is tight.

You can see a real one on this site at shaheermalik.com/llms.txt. It lists my services, the published prices, the Ship It Live sprints and these free tools, so an assistant asked about me has accurate numbers to quote.

The honest adoption picture

llms.txt is a proposal, not a standard, and nobody should sell it to you as a ranking hack. Support from the big AI crawlers is inconsistent and mostly unconfirmed: most major vendors have not said their systems read it, and Google has said its Search does not use it. Adoption has been strongest on developer documentation sites, where agents and coding tools fetch docs on demand.

Why bother, then? Because it costs half an hour, carries no downside, and helps in the cases that do exist: an agent browsing your site, a user pasting your URL into a chat, a tool built to look for it. The work also forces you to write a clear summary and decide which pages matter, which improves the rest of the site. Treat it as cheap insurance alongside the things that definitely work: fast, readable HTML, schema markup and an up-to-date sitemap.

llms.txt vs robots.txt vs sitemap.xml

File	Job	Audience
`robots.txt`	Says which paths crawlers may fetch. Permission, not meaning.	All crawlers, including AI bots like GPTBot and ClaudeBot
`sitemap.xml`	Lists every indexable URL with update dates. Completeness, not priority.	Search engines
`llms.txt`	Explains what the site is and which few pages matter, in Markdown.	Language models and AI agents
`llms-full.txt`	The full text of key pages in one file, for loading everything at once.	Models and agents working with docs

They do not overlap. Blocking an AI crawler is still done in robots.txt; listing every page is still the sitemap's job. The proposal also suggests offering Markdown versions of pages at the same URL with .md appended, which is worth doing for documentation.

What to put in yours

A summary a stranger would understand."Independent design studio in Leeds building websites for startups" beats "We craft digital experiences".
The pages that answer buying questions. Services, pricing, process, case studies, contact. Five to twenty links is plenty for most businesses.
Facts that change rarely, stated plainly. Location, who you serve, starting prices if you publish them. Keep them identical to the site.
A note on every link."Fixed prices for each package" tells a model whether to open the page.

Common mistakes are linking to every page (that is the sitemap), writing marketing copy instead of facts, using relative URLs, and forgetting to update the file when prices change. If you want a site built to be read, cited and recommended by AI assistants, that is what I do on the AI products side, and every site I build ships with structured data and an llms.txt from day one.

Common questions