If you’ve heard folks mention LLMs.txt and wondered what it is, you’re in the right place. Think of LLMs.txt as a short, human-friendly Markdown file that lives at https://yourdomain.com/llms.txt. Its job is to help language models understand your site faster by pointing them to the right docs, policies, examples, APIs, and other helpful text sources. You might hear people call it an “LLMs text file” or “LLM guide file.” Same idea. We’ll walk through what LLMs.txt is, why it matters, how to make one, where to put it, and the small details that save you from headaches later.
What LLMs.txt is, and what it’s not
LLMs.txt is a Markdown document that tells AI systems where your best reference material lives. It is not a directives file. There are no User-agent lines or Disallow rules. Instead, you write a short intro plus sections with bullet links to raw text, Markdown, or clean HTML pages that LLMs can read well. The project that popularized this approach describes it plainly as a way to help LLMs use your website content at inference time, not as a bot control mechanism. llms-txt
LLMs.txt is also not robots.txt. If your goal is to control crawling or restrict training, you still use robots.txt and bot-specific user agents, such as GPTBot or Google-Extended. LLMs.txt sits alongside those files and complements them.
Why LLMs.txt matters for your site
- It shortens the path to the good stuff. When a model lands on your domain, a clean LLMs.txt can send it straight to docs, policies, FAQs, or knowledge bases without wandering your navigation. This boosts the odds that answers pulled from your site are accurate and on brand.
- It’s easy to maintain. Because it’s Markdown, anyone on your team can edit it in a code editor or CMS and keep it versioned in Git. The spec encourages a simple structure that’s hard to mess up.
- It plays nicely with automation. Generators and scripts can compile link sections for you, then publish the file to the root of your site.
How to create your LLMs.txt file step by step
The spec is intentionally light. Follow these steps and you’ll be set.
Step 1. Outline your sections
Decide the top-level groups you want to expose. Common picks:
- Docs or Guides
- API Reference
- Policies or Legal
- Pricing or Catalog
- Examples or Tutorials
- Changelogs or Release Notes
- Optional or Archived
Each of these will become an ## H2 section with bullet links.
Step 2. Draft a title and one-line summary
At the very top of your LLMs.txt, write:
- A single
# H1with your site or product name - A one-paragraph summary as a blockquote
>right under the H1 - Optional notes as normal paragraphs or lists
That’s it for the opening. Keep it crisp and concrete. The spec calls for exactly this order.
Step 3. Write the content in Markdown, not robots syntax
This is where many articles go off track. LLMs.txt does not use User-agent rules. It’s a plain Markdown file with headings and lists. Link to source files that are text-heavy and easy to parse, such as .md or clean HTML pages. When linking, you can include a short note after the link that explains what’s inside.
Step 4. Add link sections
Create ## subheadings and use bullet lists. Each bullet is a normal Markdown link, followed by optional descriptive text. The spec and examples show exactly this pattern.
Step 5. Publish at the root
Upload the file to your site root as /llms.txt. Keep the path lowercase to match what tools expect. Your URL should be https://yourdomain.com/llms.txt.
LLMs.txt examples in Markdown, you can copy
Below are two quick examples you can adapt.
A minimal LLMs.txt for a SaaS doc site
# SEO Action Plan Help Center
> SEO Action Plan is a Google Chrome extension that helps to check the website's SEO analysis, Open graph and other issues related to website with tons of feature.
Remember:
- Pricing and API limits change. Always refer to the latest docs.
- Support replies live in our community forum.
## Docs
- [Product overview](https://docs.acme.com/overview.md): Start here
- [How billing works](https://docs.acme.com/billing.md)
- [Changelog](https://docs.acme.com/changelog.md): Latest releases
## API
- [REST reference](https://docs.acme.com/api/reference.md)
- [Auth and webhooks](https://docs.acme.com/api/auth.md)
- [Postman collection](https://assets.acme.com/postman.json): Machine readable
## Policies
- [Terms of Service](https://acme.com/legal/terms.html)
- [Privacy](https://acme.com/legal/privacy.html)
## Optional
- [Legacy API v1](https://docs.acme.com/api/v1/): Deprecated
An LLMs.txt for a content site or publication
# Example Newsroom
> We publish original reporting on technology, science, and culture. Use these links for sourcing and context.
Notes:
- Style guide applies to all sections
- Corrections are logged publicly
## Editorial
- [About and mission](https://examplenews.com/about.md)
- [Style guide](https://examplenews.com/style.md)
- [Corrections policy](https://examplenews.com/corrections.md)
## Coverage
- [Tech](https://examplenews.com/tech/index.html)
- [Science](https://examplenews.com/science/index.html)
- [Culture](https://examplenews.com/culture/index.html)
## Data
- [Methodology for surveys](https://examplenews.com/data/methodology.md)
- [Public datasets](https://examplenews.com/data/): CSV and JSON
The structure you see above mirrors the spec and real-world examples. Use H1, a one-line summary, optional notes, then H2 sections with bullet links and short descriptions.
Where to put LLMs.txt and how to test it
- Location: place the raw Markdown file at
https://yourdomain.com/llms.txt. - Accessibility: fetch it in a browser or with
curlto confirm it returns200 OK. - Content checks: scan it for dead links and ensure each important area of your site has at least one clear link.
- Automation: some teams generate the file during build and deploy it with the site. That keeps it from drifting from your docs.
LLMs.txt vs robots.txt and AI training controls
Since this trips people up, here’s the tidy summary:
- Use LLMs.txt to guide models to high-quality references on your site in Markdown form.
- Use robots.txt if you want to control crawling or opt out of training for specific bots such as GPTBot or Google-Extended. Those rely on user agent rules in robots.txt, not on LLMs.txt.
If you plan to manage training access, read up on how those crawlers respect robots.txt and how to set rules. You can add entries like User-agent: GPTBot with Disallow: / in robots.txt, and similar for Google-Extended. Keep in mind that adherence to robots.txt is voluntary across the broader web.
LLMs.txt tips, common mistakes, and a quick checklist
Tips
- Prefer links to raw Markdown, plain text, or clean HTML with minimal scripts. This keeps context light and fast to parse.
- Keep the H1 to your product or site name. Put the one-sentence hook right under it as a blockquote.
- Use clear section titles such as “Docs,” “API,” “Policies,” and “Examples.” Stick to
##for section headings.
Common mistakes
- Writing robots style directives in LLMs.txt. Those belong in robots.txt, not here.
- Burying key pages behind query parameters or heavy scripts.
- Letting it drift from your docs. Treat it like a table of contents and keep it fresh.
Checklist
- H1 set, one-line summary present
- Sections added with
## - Bullet links point to stable, text-first pages
- File published at
/llms.txt - Links checked for 200 responses
Bonus for visibility: add a Preferred Sources banner on your site
Google recently rolled out Preferred Sources for Top Stories. Your readers can star your site so it shows up more often in the Top Stories box. Google provides an official deeplink and a downloadable button you can embed. Example deeplink pattern:https://google.com/preferences/source?q=<your website URL>
Google’s announcement also links to resources for publishers. If news is part of your content mix, this is worth adding.
Tools, CMS tips, and next steps
- Where to learn the LLMs.txt format. The spec shows the expected structure with an H1 title, a blockquote summary, and H2 sections of bullet links. The examples are short and easy to mirror
- How to host it. Publish the raw Markdown at
/llms.txtso it’s reachable at a clean URL. Avoid query strings. - Keep robots.txt separate. Manage bot access and training preferences with user agent rules there, including
GPTBotandGoogle-Extended.
What to remember while creating LLMs.txt
- Keep the top of the file short and helpful.
- Use
##sections with bullet lists of links, each with a short, plain description. - Link to stable, text-first resources such as Markdown or simple HTML.
- Publish it at
/llms.txtand recheck it with every big doc or policy change. - Treat it like a living table of contents for machines and humans alike.
Want help checking your LLMs.txt and your banner?
Give our SEO Action Plan Chrome extension a try. You can use it to quickly check whether your site exposes LLMs.txt, confirm that the file resolves at the root, and verify that your Preferred Sources deeplink or button appears on key pages. It’s handy for SEOs, developers, and site owners who want a quick pass before shipping changes.
Frequently asked quick answers
Does capitalisation matter in the URL?
Use lowercase /llms.txt for the path. The term LLMs.txt is used in articles as a name, but the spec and tools expect the lowercase path.
Can I mix PDFs and HTML in my links?
You can, though plain text and Markdown are preferable. If you link to HTML, keep it clean and readable.
Do I still need robots.txt?
Yes. LLMs.txt is not a replacement for robots.txt. Use robots.txt for crawl control or training opt-outs such as GPTBot and Google-Extended.
Where can I see more examples?
The project site includes domain-specific examples so you can copy the pattern and adjust. llms-txt
By following the simple Markdown structure above, publishing your file at /llms.txt, and adding a clear Preferred Sources button, you’ll give both machines and readers a faster path to your best content. If you want a second set of eyes, fire up the SEO Action Plan extension and run a quick check across your key pages.
