llms.txt example: annotated files for non-docs sites

An llms.txt example for a SaaS product, a directory and a personal site, annotated line by line, with the v2 template and a pre-publish checklist.

Bunny Team · · 6 min read

On this page

Most llms.txt examples people cite come from developer-documentation companies, and the spec's own sample is a short mock plus one code-docs project. This post gives three annotated files for other kinds of site, the v2 template to copy, and a checklist to run before you publish. You are reading this on bunny.directory, a launch directory for startups, so weigh the tools we mention with that in mind.

An llms.txt example is a Markdown file that tells an agent what a site is and which pages matter most. It has an H1 name, a blockquote summary, H2 sections of links, and an Optional section for skippable pages. Most examples are developer docs, so this post covers a SaaS product, a directory and a personal site.

How does an llms.txt example work?

An llms.txt example gives an agent a short map first and leaves the detail behind links. The spec says the file should stay small enough to fit in context, and the detail is fetched only when needed.

The spec requires this order, which lets a parser read the file with ordinary tools:

  1. An optional byte-order mark (BOM).
  2. An H1 with the name of the project or site. This is the only required section.
  3. A blockquote with a short summary.
  4. Zero or more paragraphs or lists, with no headings, that explain the project and how to read the files.
  5. Zero or more H2 sections of file lists.
  6. An Optional section for links an agent can skip.

Each file list entry is a Markdown link, [name](url), optionally followed by a colon and a note. The links should point to LLM-friendly content, which in practice means Markdown versions of your pages.

Copy this template and fill it in:

# Site name
> One sentence on what the site is and who it serves.

Background paragraph on how to read the files. No headings here.

## Section name
- [Page title](https://example.com/page.md): What the page answers

## Optional
- [Page title](https://example.com/secondary.md): What it adds

Keep the wording concise, give every link a brief note, and avoid jargon the agent would have to guess at. The spec also suggests a test: ask an agent questions about your content, giving it only your llms.txt as a starting point, and see where it fails.

For pages with a Markdown twin, the spec proposes serving page.md or page.html.md at the same URL. Version 2 also defines two link relations: rel="alternate" type="text/markdown" points to the twin, and rel="describedby" points to the llms.txt file that covers a page. A CDN can send both as an HTTP Link header without touching any page:

Link: </docs/page.html.md>; rel="alternate"; type="text/markdown", </docs/llms.txt>; rel="describedby"

Put the file at the root to cover a whole site, or at a path such as /docs/llms.txt to cover only the pages under it. The spec says that where more than one file applies, agents should use the most specific one. The spec's live example is the FastHTML llms.txt, placed at /docs/ for that reason. For the file format in more depth, read llms.txt generator setup, file specs and search reality.

Examples

Mintlify's teardown of real files, dated 7 February 2026, analyses files from developer-focused companies. The three examples below cover other site types. The names are placeholders, so copy the structure, not the wording.

SaaS product

A SaaS file names the product in the H1, says in one sentence what it does, and sends agents to pricing and reference pages first.

# Acme Analytics
> Acme Analytics is a product analytics tool for ecommerce teams. It joins order data with ad spend to show revenue per channel.

Use the pricing page as the source of truth for plan limits. Pages under /docs/ cover setup and metric definitions.

## Start here
- [Product overview](https://example.com/product.md): What Acme Analytics reports and which data sources it reads
- [Connect a store](https://example.com/docs/connect-store.md): Setup steps for linking order and ad accounts
- [Pricing](https://example.com/pricing.md): Plans, limits and what counts as a tracked order

## Reference
- [Metric definitions](https://example.com/docs/metrics.md): How revenue per channel and refunds are calculated

## Optional
- [Changelog](https://example.com/changelog.md): Release notes by month

The summary names the product's job and its user, so an agent can judge relevance from it alone. The paragraph under the summary says which page is the source of truth for plan limits, so the agent does not guess them. Each link carries a note that says what the page answers, which lets an agent pick a link without opening it. The changelog sits under Optional, so an agent with a short context skips it.

Directory or listing site

A directory file explains what a listing holds, then links one page per region or category rather than one per listing.

# Harbor Directory
> Harbor Directory lists independent boat brokers by region. Each listing gives a broker's service area, specialties and how to make contact.

Listings are added after the broker confirms the details. Compare brokers on the region pages, which show the same fields for every listing.

## Regions
- [Pacific Northwest brokers](https://example.com/regions/pacific-northwest.md): Every listed broker in the region, with service areas
- [Great Lakes brokers](https://example.com/regions/great-lakes.md): Every listed broker in the region, with specialties

## Guides
- [How to choose a broker](https://example.com/guides/choosing-a-broker.md): What to ask before you sign anything

## Optional
- [About Harbor Directory](https://example.com/about.md): Who runs the directory and how it is funded

The spec notes that a sitemap usually lists documents that, in aggregate, are too large for a context window. A directory can hold thousands of listings, so link the category pages that group them and leave the listings to the directory's own search. The paragraph under the summary explains how listings get added, so an agent can say how much to trust one.

Personal site

A personal file answers the questions a visitor will ask about one person, so it links the CV, the work and the contact route.

# Jane Doe
> Jane Doe is a product designer. This site holds her CV, case studies and talks.

Use the CV page for roles, titles and dates. Use the case studies for outcomes, because the CV does not give results.

## Career
- [CV](https://example.com/cv.md): Roles, dates and employers in order
- [Case studies](https://example.com/work.md): Projects with the problem, the approach and the result

## Talks
- [Speaking](https://example.com/talks.md): Conference and podcast appearances, with links

## Optional
- [Reading list](https://example.com/reading.md): Books she recommends

The spec names a personal site answering questions about someone's CV as one of its intended uses. The paragraph under the summary tells the agent which page to trust for which fact, so a role's dates come from the CV and an outcome comes from a case study.

When to use it (and when not)

Publish one when your site has more pages than a reader can scan and the file takes an afternoon to write. Do not publish one expecting it to lift rankings or citations. Contentful reports no widely validated evidence that llms.txt improves AI search visibility.

No major LLM provider has committed to reading the file. Ahrefs, updated 15 June 2026, reports that OpenAI, Anthropic and Google do not currently support it. Ahrefs' study of 137,000 sites, dated 20 November 2025, found that 97% of llms.txt files are never read. John Mueller compared the file to the old keywords meta tag: a claim a site owner makes about the site, which anyone can check against the site itself.

Google's guidance points the same way. Google's AI optimization guide says site owners do not need new machine-readable files, llms.txt included, to appear in generative AI search. Chrome's Lighthouse audits sites for the file as part of its agentic browsing checks.

Checklist: will anyone read this file?

  • Host it where agents look. Wix found that 62% of live files sit at the domain root, 10% in subfolders and 28% on subdomains.
  • Run curl -I https://docs.example.com/llms.txt and confirm the server returns the file, not an error.
  • Check that every link resolves to a page with content, and prefer Markdown versions.
  • Ask an agent questions with only the file as a starting point, as the spec suggests.
  • Check your server logs for requests to /llms.txt from AI crawlers. Contentful suggests logged crawler hits as one measure, and Ahrefs found that most files get no reads at all.
  • Set a review date. GitBook advises revisiting the file after every information architecture change.

What's new (as of 3 October 2026)

Key takeaways

  • The v2 spec makes an H1 the only required section, followed by a blockquote summary and an Optional section for links an agent can skip.
  • Wix found that 62% of 1,400 live llms.txt files sit at the domain root.
  • Ahrefs' study of 137,000 sites, dated 20 November 2025, found that 97% of llms.txt files are never read.
  • Ahrefs reports, as of its 15 June 2026 update, that no major LLM provider currently supports llms.txt.
  • Google's AI optimization guide says site owners do not need llms.txt to appear in generative AI search.

FAQ

Does Google read llms.txt?

Google has not committed to crawling the file. Ahrefs reports that Google added llms.txt to its Agent2Agent protocol in April 2025 but has not committed to crawling it, and Google's AI optimization guide says site owners do not need the file to appear in generative AI search.

Where should the llms.txt file go?

Put it at the site root, /llms.txt, to cover the whole site, or at a subpath such as /docs/llms.txt to cover only the pages under that path. The spec says agents should use the most specific file where more than one applies. Wix found that 62% of live files sit at the root.

Do I have to write llms.txt by hand?

No. The spec lists Mintlify, GitBook, Yoast SEO, AIOSEO and Wix as platforms that can generate the file. GitBook still advises reviewing the output and tightening it around your canonical content, because generated files usually need editing to get better results.

To draft your own file, start with the llms.txt generator on bunny.directory and then check the result against the template above.

New guides by email

Keep reading