What Is llms.txt, and Does Your Website Actually Need One?

Key takeaways

• llms.txt is a proposal, not a standard. Jeremy Howard published it in September 2024, and version 2 was last modified on 10 August 2026. Only the H1 title is required.
• It will not help you rank in Google. Google's AI optimization guide says doing it will "neither harm nor help" visibility in Google Search, as Google Search ignores the files.
• Most files sit unread. Ahrefs studied 137,210 domains in May 2026. About 38,000 had a file, and 97% of those saw no requests for it that month.
• It has a real use with coding tools. The proposal itself says it is used most for software documentation, where agents follow it to reach API references. Docs sites have the strongest case.
• For a small business site, treat it as optional. It takes under an hour to write and costs nothing to host, but do it after the basics. Do not pay for one.
• Chrome's check does not change that. Lighthouse has an llms.txt audit, and a missing file is marked not applicable, not a failure.

The short answer

llms.txt is a plain text file in Markdown, usually at yoursite.com/llms.txt. It gives AI agents a short summary of your site and a curated list of links. It is not a ranking factor, and no major AI company we read has said its crawlers rely on it. If you run developer docs, publish one. If you run a service business and your site basics are in good order, it is a cheap and harmless experiment. If your pages are not indexed yet, fix that first.

What the file does

An agent reading a web page has to dig through navigation, scripts and layout to reach the text. Howard's proposal argues that agents do better with concise information gathered in one place, because they can only hold so much at once. The file offers that place, and detail sits behind its links.

It is often called "robots.txt for AI", and that is misleading. These three files do different jobs:

So llms.txt cannot block anything. If you want to control AI crawlers, that is a robots.txt job. The proposal also expects the file to be used on demand while an agent helps someone, more than for model training.

What v2 changed

Most guides on page one still describe the first version. Howard's changes page sets out the differences, and they matter if you plan to do this properly.

The discovery part is the new idea. A page can point to its Markdown copy with rel="alternate" type="text/markdown", and to the covering file with rel="describedby". A small business does not need this. A docs team may well want it.

The format, in one example

The order is fixed: an H1 name, a blockquote summary, optional notes with no headings, then H2 sections holding link lists. Only the H1 is required. A section called "Optional" holds links an agent can skip when it needs less.

Here is a template for a service business. It is an illustration, so swap in your own details.

``` # Example Plumbing Co

Licensed plumber serving Denver. Emergency callouts, repairs and bathroom fitting for homes and small businesses.

Call for emergencies. Quote requests are answered within one working day.

Services

• Emergency plumbing: What counts as an emergency and how callouts work
• Bathroom fitting: Process and what a quote includes

Company

• About us: Licences, insurance and service area
• Contact: Phone number, form and hours

Optional

• Blog: Guides on common plumbing problems

```

The proposal gives a few rules for writing it: use clear language, add a brief note to each link, avoid unexplained jargon, and test the file by asking an agent questions with only that file to go on. It says links should point to LLM-friendly content, such as Markdown versions of pages. Plain web pages still work as links, though an agent then has to deal with the HTML.

A real example: ours

We publish our own llms.txt at the site root. It is about 6 KB with 39 links. The sections are Main pages, Services, How projects work, Guides on the October 2026 PHP security flaws, Guides on cost and platform choices, Free tools, and All posts.

Here is the start of it, cut for space:

``` # Amariq

Amariq is a design and development studio that redesigns underperforming websites for marketing teams and ecommerce brands, building on Webflow, WordPress, Shopify and Framer. Projects are priced from scope, never by the hour.

How projects work

• Projects are priced from scope, not by the hour.
• Payment is 40% to start, 30% at design approval and 30% before launch.

```

Two choices are worth copying. We put facts that people often ask about in the file itself, such as how projects are paid for, so an agent need not hunt for them. And we group posts by topic and do not paste the whole sitemap. We make no claim that the file has earned us anything, since that would be hard to prove, and the evidence below gives no reason to expect it.

What the evidence says

These are the figures from people who have measured it. Each measures something slightly different, so read the limits.

The picture is fairly clear. Adoption is growing quickly and use is low, but not zero. Fetches come largely from bots, and the file sees real use mostly in coding tools.

What Google says, and why it looks mixed

Google has two messages, because different teams have different jobs.

Google Search says no. Its AI optimization guide, last updated 10 July 2026, says you do not need new machine readable files, AI text files, markup or Markdown to appear in Search, as Search does not use them. It adds that it is "completely fine" to keep llms.txt files for other services that use them. Google's AI features page says the same about AI Overviews and AI Mode.

Chrome's Lighthouse has a check. It sits under Agentic browsing audits, and the page was last updated on 5 May 2026. A server error when fetching the file fails the audit. A 404 is marked not applicable, because the file is optional. So this audit does not mean Google wants one from you, and an empty result is nothing to fix. A tool that warns you about a missing llms.txt is going further than Chrome does.

So an SEO seller who says llms.txt lifts rankings is contradicting Google's own page. Ask for their proof.

Who should publish one?

Better work usually comes first: clear pages, plain answers, real facts and pages that load. Agents and search systems read those too. Our website audit checklist is a good place to start.

How to add one in 30 minutes

1. Pick 10 to 20 pages. Choose the pages you would send a new customer. Leave out tag pages and anything thin.
2. Write the summary. Say what you do, for whom, and where, in one or two sentences a stranger could understand.
3. Write a note for each link. Say what the page answers, not what it is called.
4. Save it as llms.txt and upload it to your site root, then open the address in a browser to see it load as plain text.
5. Test it. Paste it into an AI assistant and ask three questions a customer would ask, as the proposal suggests.
6. Add a yearly reminder to review it. A stale file with broken links or old prices is worse than none.

Do not put in anything you would not want published. It is a public file.

Generators and platform support

You may not need to write it at all. The proposal lists the first four rows below, and Ahrefs lists its own tool on its site. These are the makers' descriptions:

On Framer, its docs say you add the file yourself. On a static site you just upload it. Whichever you use, read what a generator produces, because a list of every URL is not the curated guide the format is meant to be. A generator can draft the structure, but only you know which ten pages matter most. There is no reason to pay for this.

How we help at Amariq

We build sites on Webflow, WordPress, Shopify and Framer, and on any new build we can add an accurate llms.txt with little fuss. The bigger job is making the pages themselves clear enough to quote, with structure and copy that answer real questions. See our services to find out how we work.

Frequently asked questions

What is llms.txt in simple terms?

It is a short Markdown file on your site that tells AI agents what your site is and which pages matter most. Think of it as a guide, not a rule, since it does not block or allow anything.

Does llms.txt help SEO?

Not in Google. Its guide says the files will neither harm nor help visibility in Search, as Search ignores them. We found no proof from other studies of a boost in AI answers either.

Do ChatGPT and Claude read llms.txt?

Ahrefs recorded fetches from GPTBot and Claude-Code on the small share of files that got traffic. A fetch is not proof of use, and the pages we read found no AI company stating its crawlers rely on it.

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

robots.txt says what crawlers may access. llms.txt offers a short guide to useful content. One sets access, the other suggests where to look, so llms.txt cannot replace robots.txt.

Where does the file go?

Usually at the root, as yoursite.com/llms.txt. Version 2 also allows subpaths, such as /docs/llms.txt, which covers just the pages beneath it. The most specific file applies.

Is there a free llms.txt generator?

Yes. Ahrefs has a free one, Yoast and AIOSEO offer it in WordPress, and Wix is listed as generating one by itself. You can also write it by hand in any text editor in under an hour.

What is llms-full.txt?

It is a name you will see for a single file holding a site's full documentation, and some docs tools produce one. The main v2 proposal text we read does not define it, so treat it as a tool convention.

Can llms.txt hurt my site?

Google says it will not harm your visibility in Search. The real risks are out-of-date facts and private details in a public file, so keep it accurate and short.