llms.txt is a plain-text file, written in Markdown, that sits at yourdomain.co.nz/llms.txt and gives AI systems a short, clean summary of your website with links to the pages that matter most. It was proposed by Jeremy Howard in September 2024. Google has said in writing that it does not use the file, and none of the major AI companies has committed to reading it, so treat it as cheap insurance rather than a ranking tactic: worth an afternoon, not worth a retainer. We publish one for our own site, and this guide explains what it is, who actually reads it, what matters more, and how to write one.
What is llms.txt, exactly?
Think of it as a cover letter for your website, addressed to machines. Web pages are built for people: the useful text is wrapped in navigation menus, cookie banners, scripts and adverts. An AI assistant that wants to know what your business does has to fetch all of that and pick out the meaning. The proposal describes the problem in its own words: "an HTML page wraps its information in navigation, ads, and JavaScript, and converting it back into clean text is difficult and imprecise."
The file itself is deliberately simple. According to the specification, it contains, in order:
- A single top-level heading with the site or business name. This is the only required part.
- A short summary in a blockquote: what the business is, who it is for, the key facts.
- Optional paragraphs or lists with more detail.
- Sections under second-level headings, each a list of links to important pages with a one-line note about each.
- By convention, a section called "Optional" for links an assistant can skip when it is short on room.
It lives at the root of the domain, the same place as robots.txt, and a person can read it in a browser. That is the whole thing.
Where did llms.txt come from?
The proposal was published on 3 September 2024 by Jeremy Howard, co-founder of Answer.AI and fast.ai, and the specification was revised in August 2026. It grew out of a practical annoyance: developers were pasting whole documentation sites into AI assistants and running out of room, because an assistant can only hold so much text in view at once. A single curated file solves that for the people who use it.
Adoption so far has been strongest among technology companies with large documentation sites. Anthropic publishes one for its developer documentation, and Cloudflare publishes one that links to a separate file per product. At the other end of the scale, we publish one for a small Auckland web design business, which we use as the worked example below.
Does Google use llms.txt?
No, and it has said so plainly. Google's guide for site owners on its AI features states: "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn't use them." The same page adds that it is "completely fine" to create and maintain llms.txt files for other services that use them.
So for Google Search, AI Overviews and AI Mode, an llms.txt file does nothing, good or bad. What gets you into Google's AI answers is what has always got you into Google: crawlable pages, helpful content and the ranking systems it already runs. We covered that in Is SEO still worth it?
Do ChatGPT, Claude or Perplexity read llms.txt?
This is where the honest answer is "not officially". Each company publishes documentation about how its systems fetch web pages, and none of it commits to reading llms.txt:
- OpenAI documents OAI-SearchBot, which is "used to surface websites in search results in ChatGPT's search features", alongside GPTBot for training and ChatGPT-User for actions a user asks for directly.
- Anthropic documents Claude-SearchBot, Claude-User and ClaudeBot, for search results, user-requested fetches and training respectively.
- Perplexity documents PerplexityBot, which it says is "designed to surface and link websites in search results on Perplexity", and Perplexity-User for user-initiated visits.
All three describe crawling ordinary web pages. If an assistant is asked about your business, it will fetch your actual pages, and it can fetch llms.txt if it happens to look for it or a user points it there. That makes the file a low-cost hedge with an uncertain payoff, which is exactly how our earlier guides to GEO and to getting recommended by AI describe it. If anyone tells you llms.txt is a proven way to get recommended by ChatGPT, ask them for the documentation. There is not any.
What matters more than llms.txt for AI visibility?
If you have one afternoon for "AI search", spend it on these before the file, in this order.
1. Do not block the AI crawlers
Every search crawler listed above obeys robots.txt; the user-initiated fetchers are the exception, by design. A surprising number of sites block them by default because a plugin or a hosting provider decided to, and then no llms.txt in the world will help, because the assistant cannot read anything. Open yourdomain.co.nz/robots.txt and check. Ours, at godigital.co.nz/robots.txt, names each of them and says welcome.
2. Write pages that answer questions directly
Assistants extract passages. A service page that opens with what the service is, who it is for, where you offer it and what it costs gives them something to quote. A page that opens with "welcome to our website" does not.
3. Publish the facts an assistant needs to trust you
Who you legally are, where you are, what you charge, and what happens if a customer wants to leave. We publish ours as a plain-language answers page and a pricing file in Markdown, partly for buyers doing due diligence and partly because those are the questions assistants get asked about us.
4. Keep your details consistent everywhere
Assistants cross-check your site against your Google Business Profile, reviews and directories. One name, one phone number, one address format, everywhere.
5. Add structured data
Organisation, LocalBusiness, Service and FAQ markup describe your business in a format machines already parse. Unlike llms.txt this is well established, and it helps ordinary search as well.
Then, if you still have half an hour, write the file.
How do you write an llms.txt file?
Use our own file as the pattern. It is generated from the same data as our web pages, so it cannot drift out of date, and it contains, in order:
- The business name as the heading, and a two-sentence summary: what we do, for whom, from what price.
- The legal entity: our registered company name, NZBN and Companies Register link, plus a note that an unrelated company with a similar name is registered in the same suburb. Assistants confuse the two otherwise.
- Contact details and service area.
- Links to the pages that answer the most common questions: pricing, answers, services, the audit tool.
- Every service with a one-line description and its URL.
- Full pricing, including what is deliberately not included.
- Every blog post, location page and industry page, as a plain list.
Yours can be shorter. For a typical NZ service business, a heading, a three-line summary, contact details, a list of services and a link to your pricing or booking page is plenty. Write it in a text editor, save it as llms.txt, upload it to the root of the site, and check it loads in a browser. If your site is built on a platform that does not let you add files at the root, ask your provider; if they cannot do it, that tells you something about the platform.
Three rules: write it in plain sentences a stranger would understand; only state things that are also true on your pages; and update it whenever those pages change.
What are the common llms.txt mistakes?
- Publishing llms.txt while blocking the crawlers. The file is pointless if robots.txt turns the reader away at the door.
- Letting it go stale. An old price in llms.txt is a wrong answer waiting to be repeated. If you cannot generate it from your site data, put a calendar reminder on it.
- Filling it with marketing. "Award-winning, industry-leading solutions" tells an assistant nothing. "Emergency plumber, North Shore, callout fee listed, 24 hours" tells it everything.
- Expecting Google results. Google does not read it. Any effect will come from other assistants, and it will be modest.
- Paying a retainer for it. It is a text file. Be wary of anyone billing significantly for "llms.txt optimisation".
Do this today
- Open yourdomain.co.nz/robots.txt and check that none of the AI crawlers above are blocked
- Ask ChatGPT or Claude what your business does and what it charges, and note what it gets wrong
- Fix the source of each wrong answer on your website first
- Write a ten-line llms.txt in a text editor and upload it to the root of your site
- Check it loads in a browser, and check it again when your prices change
Where we come in
Every site we build ships with this layer as standard: crawlers welcomed, structured data on every page, service pages written answer-first, and an llms.txt generated from the site's own content so nothing has to be maintained by hand. It is part of what "set up to be found in AI answers" means on our plans, from the $49 a month Starter upward. If you have an existing site and want to know whether assistants can read it, our free website audit includes AI-readiness checks alongside the usual SEO, speed and security ones.
Frequently Asked Questions
Is llms.txt the same as robots.txt?
No. Both live at the root of your domain, but robots.txt tells crawlers which pages they may or may not visit, while llms.txt is a plain-language summary of your site for AI systems that choose to read it. Robots.txt is a long-established standard that every major crawler obeys; llms.txt is a 2024 proposal with no such commitment behind it.
Does llms.txt improve Google rankings?
No. Google states that you do not need to create AI text files, markup or Markdown to appear in Google Search, including its AI features, because Google Search does not use them. It also says creating such files for other systems is fine and will not hurt you. Your Google visibility still comes from crawlable pages, helpful content and normal SEO.
Will llms.txt get my business recommended by ChatGPT?
Not by itself. OpenAI, Anthropic and Perplexity all document crawlers that read ordinary web pages, and none commits to reading llms.txt. What gets a business recommended is being readable, accurate and consistent: crawlers allowed, pages that answer questions directly, published pricing and identity, and a complete Google Business Profile with genuine reviews.
How do I check if a website has an llms.txt file?
Type the domain into your browser followed by /llms.txt, for example godigital.co.nz/llms.txt. If a plain text page appears, the site has one. If you get a page not found error, it does not. The file is meant to be readable by people as well as machines, so it should make sense when you read it.
Should I pay someone to create an llms.txt file?
It should not cost much. The file is a few dozen lines of plain text that a business owner can write in an afternoon, and any provider building a modern website should include it as a matter of course, as we do on every plan. Be wary of anyone selling it as a standalone service at a significant price or promising it will get you into AI answers.
The bottom line
llms.txt is a small, sensible, unproven idea. Google ignores it, other assistants might read it, and it costs almost nothing to publish, so publish it, after you have made sure the crawlers can reach your site and your pages say the right things. If you want to see whether your site is ready for AI assistants, run the free website audit, or read our own llms.txt to see what one looks like in practice.