In classic SEO, `sitemap.xml` told search engines where your pages live. In Generative Engine Optimization, `llms.txt` tells AI models what your pages mean. It is a plain Markdown file at the root of your domain that hands ChatGPT, Claude, Perplexity, and other AI agents a curated, token-efficient summary of your most important content, so they can describe and cite you accurately instead of guessing.
This is the hands-on build guide. By the end you will have written a real `llms.txt`, deployed it on your platform, and confirmed it is live and correct. If you want the strategy behind the format, which pages to prioritize and why it moves AI visibility, read [how to create an llms.txt file for GEO](/blog/how-to-create-llms-txt-file-for-geo) alongside this. Here we focus on getting the file built and shipped.
Key takeaways
- `llms.txt` is a single Markdown file served at `https://yourdomain.com/llms.txt`. There is no XML, no special encoding, just headings, a summary, and curated links with one-line descriptions. - The structure is fixed enough to be easy: an H1 with your name, a blockquote summary, then H2 sections grouping your most important links, each link followed by a short description. - Curate ruthlessly. The value is in choosing your best pages and describing them, not in listing everything. A bloated file defeats the point. - Deployment differs by platform. On a static or code-based site you drop the file in the public root; on Shopify or WordPress you route the path or use a plugin or app. - Verify by loading the URL in a browser, confirming it returns plain text with a 200 status, and checking the links resolve.
Step 1: Choose the pages worth including
Before writing anything, list the pages an AI actually needs to understand and recommend you. Think about what a model would need to answer "what is [your brand] and should I use it." That usually means your homepage or About page, your core product or feature pages, pricing, key documentation, and a few cornerstone guides. Leave out login screens, cart pages, tag archives, and thin or duplicate pages. The whole point of `llms.txt` is curation, so a focused list of ten to twenty strong pages beats a dump of two hundred.
Step 2: Write the file in Markdown
Create a file named `llms.txt` in a plain text editor. The format is standard Markdown with a specific shape: one H1 for your project name, a blockquote one-liner summarizing what you do, then H2 sections that group links, each with a short description after a colon.
Here is a working example you can adapt:
``` # Acme Analytics
Acme Analytics is a GEO monitoring platform for DTC brands that tracks how AI models mention and cite them.
## Core pages - [About Acme](https://acme.com/about): Who we are, what we build, and who it is for. - [Pricing](https://acme.com/pricing): Current plans and what each tier includes. - [Product overview](https://acme.com/product): How the platform works, end to end.
