The short answer
A good llms.txt file has a clear title, a one-sentence summary, and a short list of your most useful pages, each with a one-line description. What you list depends on your site type: software points to docs and pricing, content sites to their best guides, stores to category pages and policies, and local businesses to services and contact details. It is a proposal, not a standard, and it does not allow or block any crawler.
Last updated: October 7, 2026. This post describes the llms.txt proposal as of version 2 (August 2026).
Below are five complete llms.txt examples you can adapt, plus a blank skeleton to start from. All names and URLs are made up.
llms.txt is a proposal, not an official standard, and it does not allow or block any crawler. Robots.txt still controls access. Only the title is required, and everything else is convention. I publish one on my own sites because it costs little, not because it is promised to lift rankings or earn citations.
Before you start: new to the format? I'd read What is llms.txt, and do you need one. Unsure how it differs from your other files? I compare them in robots.txt, sitemap.xml and llms.txt: what each file does. Want a step-by-step walkthrough? Use How to write an llms.txt file, with template.
A blank skeleton
Copy this, replace everything in brackets, and delete any section you do not need.
# [Site or business name]
> [One sentence: what the site is and who it is for.]
[Extra context, if useful: one or two lines, such as currency or coverage area.]
## [Main section, e.g. Docs, Shop or Services]
- [Page title](https://yourdomain.com/page): What the reader will find there
- [Page title](https://yourdomain.com/page): What the reader will find there
## [Second section, e.g. Policies or Contact]
- [Page title](https://yourdomain.com/page): What the reader will find there
## More
- [Page title](https://yourdomain.com/page): A lower-priority or older page
Example 1: A software product
For SaaS and developer tools, the file is most useful when it points to documentation, because that is where AI coding tools look for answers.
# Ledgerly
> Ledgerly is invoicing and expense software for freelancers and small agencies.
Docs cover the web app and the REST API. Pricing is in USD.
## Getting started
- [Quickstart](https://ledgerly.example/docs/quickstart): Create an account and send your first invoice
- [Importing data](https://ledgerly.example/docs/import): Bring in clients and past invoices from a CSV
## API
- [Authentication](https://ledgerly.example/docs/api/auth): API keys and token scopes
- [Invoices endpoint](https://ledgerly.example/docs/api/invoices): Create, update and list invoices
- [Webhooks](https://ledgerly.example/docs/api/webhooks): Events and payload formats
## Company
- [Pricing](https://ledgerly.example/pricing): Plans, limits and what each includes
- [Security](https://ledgerly.example/security): How data is stored and protected
## More
- [Changelog](https://ledgerly.example/changelog): Release history
When I write the descriptions, I say what the reader gets, not just what the page is called.
Example 2: A blog or content site
A publisher should not list every post. Pick the evergreen guides that best represent what you know.
# Trailhead Notes
> Practical guides to day hikes and backpacking trips in the Pacific Northwest, written by working guides.
## Start here
- [How to plan your first overnight hike](https://trailheadnotes.example/first-overnight): Gear, route choice and a packing checklist
- [Reading a trail map](https://trailheadnotes.example/trail-maps): Contour lines, scale and common symbols
## Safety
- [What to do if you meet a bear](https://trailheadnotes.example/bears): Steps by species, with a printable card
- [Water treatment compared](https://trailheadnotes.example/water): Filters, tablets and boiling
## More
- [Gear reviews archive](https://trailheadnotes.example/reviews): Older reviews, some products discontinued
I put older or secondary pages in a lower-priority section like "More." You may also see it named "Optional," but that name has no special meaning in version 2, so I call mine whatever suits the site.
Example 3: An online store
Stores have thousands of product pages, and listing them would defeat the purpose. Link the pages that explain how the store works.
# Birchwood Goods
> Birchwood Goods sells handmade wooden kitchenware, shipped from Portland, Oregon.
Prices are in USD. Orders ship within three business days.
## Shop
- [All categories](https://birchwoodgoods.example/shop): Cutting boards, utensils and serving ware
- [Care guide](https://birchwoodgoods.example/care): How to clean and oil wooden items
## Policies
- [Shipping](https://birchwoodgoods.example/shipping): Rates, delivery times and countries served
- [Returns](https://birchwoodgoods.example/returns): 30-day returns and how to start one
## About
- [Our makers](https://birchwoodgoods.example/makers): Who makes the products and where
I always include policies for a store, because those are the questions assistants get asked most.
Example 4: A local business
Keep it short and factual. Include what a person would ask before visiting.
# Harbor Dental
> Family dental practice in Duluth, Minnesota, accepting new patients.
## Services
- [Services overview](https://harbordental.example/services): Cleanings, fillings, crowns and emergency care
- [New patient guide](https://harbordental.example/new-patients): What to bring and what to expect at a first visit
## Practical
- [Hours and location](https://harbordental.example/contact): Address, parking and opening hours
- [Insurance](https://harbordental.example/insurance): Plans we accept and payment options
I only include details that are true and kept current. An outdated address or hours listing is worse than leaving the page out.
If your site is a collection of tools, I describe what each does and what it does not do. That second half stops an assistant from sending someone to a tool that cannot answer their question.
# Example Checker
> Free web tools that check how a website looks to search engines and AI assistants, with no sign-up needed.
Each tool checks one URL or domain at a time and reports what it finds. None of them change your site.
## Tools
- [SEO checker](https://examplechecker.example/tools/seo): Checks on-page basics for one URL; does not crawl the whole site or track rankings
- [AI crawler checker](https://examplechecker.example/tools/ai-crawlers): Shows which AI crawlers robots.txt allows; does not edit your robots.txt
- [llms.txt checker](https://examplechecker.example/tools/llms-txt): Confirms the file exists, is plain text and starts with a title; does not judge the quality of your descriptions
## Guides
- [Allow or block GPTBot](https://examplechecker.example/guides/gptbot): How to edit robots.txt for OpenAI's crawler
- [Reading your results](https://examplechecker.example/guides/results): What each pass, warning and fail means
## More
- [How the checks work](https://examplechecker.example/about/method): What each tool tests and its known limits
- [Release notes](https://examplechecker.example/changelog): New checks and fixes, with dates
Patterns across good examples
- A plain summary. One or two sentences that say what the site is and who it is for.
- A short list. Roughly 10 to 30 links, not a sitemap.
- Real descriptions. Each link says what the reader will find, and for tools, what it does not do.
- Absolute URLs. Full addresses starting with
https://.
- Pages that answer questions. Docs, pricing, policies, guides and contact details.
- A separate section for lower-priority links, if you have any. Examples 1, 2 and 5 include one; Examples 3 and 4 do not need one.
What to leave out
- Pages behind a login or paywall that a tool cannot read
- Duplicate or near-duplicate pages
- Thin pages with little content
- Outdated pages and anything you plan to remove
- Tracking parameters in URLs
- Marketing slogans in place of a summary
What it does not do
An llms.txt file does not grant or deny access. If you want to allow or block AI crawlers, edit robots.txt, and use the AI crawler checker to see what your current rules do. I would not expect the file to lift your rankings, and I would not expect it to guarantee that an AI answer cites you. Support varies from tool to tool. To see what large sites actually publish, read my AI-readiness scan of 97 large sites.
Build and check yours
I start from the example closest to the site, or from the blank skeleton above, replace the details, and trim until every link earns its place. Then I upload it to the site root so it loads at yourdomain.com/llms.txt, and run the domain through the free llms.txt checker. It confirms the file exists, is not empty, is plain text, and starts with the title heading the format expects.
Related posts
AI coding tools look to documentation for answers, so docs are what the file should surface.
What we did: Point to quickstart, API reference, pricing and security pages, with descriptions that say what the reader gets.
Listing every post defeats the purpose of a short, curated file.
What we did: Pick the evergreen guides that best represent what you know and put older pages in a lower-priority section.
Product pages number in the thousands, while assistants most often get asked about policies.
What we did: Link category pages, care guides, shipping, returns and an about page rather than individual products.
People ask practical questions before visiting, and an outdated address or hours listing is worse than leaving the page out.
What we did: Keep it short and factual, covering services, a new patient guide, hours and location, and insurance.
Without limits stated, an assistant may send someone to a tool that cannot answer their question.
What we did: Describe what each tool does and what it does not do, then add guides and a page on how the checks work.
Rankings and recommendations reflect our own testing and criteria. Pricing and features change often — always confirm current details directly with the provider.
Comments
No comments yet — be the first.