What is llms.txt?
llms.txt is a plain text file written in Markdown and placed at the root of a website. It gives AI agents and tools a short summary of the site and links to its most useful pages. It is a proposal from 2024, not an official standard. Google says Google Search does not use it, but other services may, and it takes little time to add.
How to write and publish llms.txt in 30 minutes
See what you have now
Run the check above. Its readiness items show whether /llms.txt answers today, beside crawler access, business schema and text readable before scripts run.
Start with the one required line
The file opens with an H1 holding your business or project name, such as
# Example Dental Clinic Dubai. The proposal says this is the only required section.Add the summary
Under the H1, write a blockquote of one to three sentences: what you do, where and for whom. It is the part an agent reads first.
Add short notes if they help
Paragraphs or lists without headings can follow, for facts an agent should know, such as languages, emirates served or licence details.
List key pages under H2 headings
Each H2 holds a Markdown list of links, each with a short note. Group them, for example Services, Locations, Guides and Contact.
Put extras under Optional
An H2 named Optional holds secondary links that an agent can skip when it needs a shorter context.
Publish it at the root
Save it as UTF-8 plain text so it opens at
https://yourdomain.ae/llms.txt. The address must return the file itself, not an HTML page or a redirect to the homepage.Check it, then keep it current
Run the next check on the full AI visibility checker. Update the file whenever you add or retire a service, a branch or a key page.
The llms.txt format, part by part
Five parts in a fixed order; only the first is required.
| Part | Rule | Example |
|---|---|---|
| H1 | Required. The site or project name, one line | # Example Dental Clinic Dubai |
| Summary | A blockquote with the key facts needed to understand the rest | > A DHA licensed dental clinic in Jumeirah, Dubai, treating adults and children in English and Arabic. |
| Notes | Paragraphs or lists, with no headings | Open Saturday to Thursday. Emergency appointments by phone. |
| File lists | H2 sections, each a Markdown list of links | ## Services |
| Link line | A dash, the page name in square brackets, the full address in round brackets with no space between, then a colon and a note | - [Dental implants] then (https://yourdomain.ae/implants) then : treatment steps and aftercare |
| Optional | An H2 named Optional for links an agent may skip | ## Optional |
Format as set out at llmstxt.org. The example clinic and its details are an illustration.
What llms.txt does, and what it does not
What it does. It gives an AI agent, or a tool that reads your site when someone asks it to, a short and clean map: who you are and which pages matter, without menus, scripts or cookie banners in the way. The proposal also suggests offering clean Markdown versions of important pages at the same address with .md added. Growth GPS does this for every guide, this one included, and links each Markdown twin from the page head.
What it does not do. It is not robots.txt. It neither allows nor blocks any crawler: access rules still live in /robots.txt, which RFC 9309 places at the top level of the site and describes as rules that are not a form of access authorisation. Nor does it help in Google Search. Google's guidance says Google Search ignores llms.txt, and that the file will neither help nor harm visibility or rankings there.
Why add it anyway. It takes half an hour, other services may read it, and writing it forces a useful discipline: saying in three sentences what you do, and choosing the pages that prove it. That is why the AI visibility checker gives it 15 of its 100 readiness points, well below crawler access, which carries 40.
Before you publish llms.txt
- The file opens at /llms.txt on your main domain and returns status 200
- It is plain UTF-8 text, not an HTML page or a redirect to the homepage
- The first line is an H1 with your business name
- A blockquote summary says what you do, where and for whom
- Every link is a full https address that opens without a login
- Links point to pages with real text, not only to PDFs or image galleries
- Arabic pages are listed too, with their own short notes
- Secondary links sit under an H2 named Optional
- robots.txt still allows the AI crawlers you want, because llms.txt does not change access
- Facts in the file match your website and your Google Business Profile
- Someone owns the file and updates it when services or branches change
Who uses this guide
Web developers
Add the file in one deploy and put it on the handover checklist next to robots.txt and the sitemap.
Documentation and software teams
Give tools that read docs on request a short route to the pages that matter, with clean Markdown copies of long pages.
Clinics, schools and hotels
List licence, branch, admissions or room pages, so an agent that reads the file reaches the right page first.
Online stores
List category pages, delivery and returns policies and the size guide, rather than thousands of product links.
Marketers
Use the summary as a test of positioning: if it cannot be said in three sentences, the website will not say it clearly either.
IT leads
Add the file to the release process, so it is versioned, reviewed and never left pointing at retired pages.
Worked examples
A report that says No llms.txt
Suppose the readiness items show No llms.txt while the other three items pass. Publish a short file this week. It is the smallest of the four items, so do it after anything that blocks crawlers, never instead of it.
A file that is really the homepage
Suppose /llms.txt opens your homepage because the server sends unknown addresses there. The checker still reports no llms.txt, because it accepts plain text only. Upload the file to the web root and confirm that the address returns the text itself.
What changes in 2027 for llms.txt
Now
- llms.txt is still a proposal, first published in 2024 at llmstxt.org, with one required section: the H1 name.
- Google says Google Search ignores llms.txt and other AI text files, and that keeping them for other services is fine (2026).
- Crawler access is still set in robots.txt, the Robots Exclusion Protocol written up as RFC 9309.
Next
- AdEngage UAE expects more AI agents in 2027 to read websites on a person's behalf, which makes a short, honest map of the site more useful than a long one.
- Watch the official pages of the AI providers that matter to you; if one documents support for llms.txt, tune the file for that service.
- Keep the file small and current. A file that sends agents to retired pages misleads them, so review it with every release.