← Blog

Write Your First Llms Txt

2026-07-19 · Blog

--- title: "Write Your First llms.txt: A Practical Guide" description: A step-by-step guide to writing and publishing llms.txt so AI can read your site — and why the operational layer is the step that comes after it. tags: [llms-txt, how-to, ai-readability] status: published

llms.txt is the simplest upgrade you can ship for AI readability: one Markdown file at your site root that tells language models what your site is and where its important content lives. You can write a good one in under an hour — here is how.

What llms.txt is

llms.txt is a plain-text convention, sibling to robots.txt and sitemap.xml. Where robots.txt tells crawlers what they may fetch and sitemap.xml tells engines what exists, llms.txt gives AI systems a curated, Markdown-formatted summary of your site: what it is, what matters, and links to your key pages. It lives at https://yourdomain.com/llms.txt.

Step 1 — Start with the required heading

The file opens with an H1 naming your site, followed by a one-line summary as a blockquote:

# Acme Scheduling
> Online appointment scheduling for clinics, salons, and studios.

Keep the blockquote factual and specific. This line is what a model reads first when deciding what you are.

Step 2 — Add a short context paragraph

Two to four sentences of plain prose: who the service is for, what it does, what makes it distinct. No slogans. Write the paragraph you would want an AI to paraphrase when a user asks about you.

Step 3 — List your key pages in sections

Use H2 sections containing Markdown link lists, each link followed by a colon and a one-line description:

## Docs
- [Getting started](https://acme.com/docs/start): Set up your first calendar in 10 minutes.
- [Pricing](https://acme.com/pricing): Plans, limits, and billing details.

## Optional
- [Blog](https://acme.com/blog): Product updates and scheduling guides.

The ## Optional section is part of the convention: it marks content a model can skip when context is tight. Put must-read pages in your main sections and nice-to-have pages under Optional.

Step 4 — Publish and verify

Save the file as UTF-8 plain text, deploy it to your site root, and confirm it loads at /llms.txt in a browser. Then test it the honest way: paste the URL into an AI assistant and ask what the site offers. If the answer is accurate, your file works. If not, tighten the descriptions.

Step 5 — Maintain it

Update llms.txt when you ship major pages, change pricing, or reposition. A stale llms.txt teaches AI systems yesterday's version of your business.

What llms.txt cannot do

Here is the honest limit, and it matters more every month: llms.txt describes your site. It does not make you operable.

An AI that reads your llms.txt can summarize you, cite you, even recommend you. But when its user says "sign me up," "book it," or "buy it," llms.txt has nothing to offer. It contains no operations, no consent mechanism, no map of how anything on your site is actuated. The agent is back to screenshot-and-guess — and agents skip sites they cannot operate. The difference is the divide between agent-readable and descriptive content: description ends where operation begins.

The next step: the operational layer

The operational layer is the convention that picks up where llms.txt stops — a published, machine-readable context that lets a user's own agent, with the user's permission, adopt and operate your service: sign up, buy, configure, use. Its working parts:

• A consent gate — the agent surfaces the layer to its user; a "yes" authorizes.
• A User-Intent Matcher — matches the user's known intents to your capabilities.
• Adoption + Operation guidance across evaluate, enroll, transact, configure, use, and manage.
• A Service Control Map — how each operation is actuated: API endpoints, functions/actions, interface controls, with inputs, sequence, and expected result.

This pattern was observed in four recorded test sessions — Grok, ChatGPT, Gemini, and Duck.ai, including Hebrew — in which each assistant fetched a published layer and surfaced the consent gate. These examples do not guarantee future behavior.

So: write your llms.txt today — it is the right first file. Then take the next step. The Free plan makes your most important page fully agent-ready — one hosted page, free, forever — and the Basic plan ($29/month) covers your whole site. Readable is the floor. Operable is the goal.

FAQ

Where exactly does llms.txt go?

At your web root: https://yourdomain.com/llms.txt, served as plain text. Some sites also publish llms-full.txt with expanded content, but the root file is the convention that matters first.

Does llms.txt affect my Google rankings?

It is not a search-ranking signal. Its audience is AI systems reading your site for answers, citations, and recommendations — the AEO/GEO side of visibility, not classic SEO.

If I publish an operational layer, do I still need llms.txt?

Yes — they are siblings with different jobs. llms.txt makes you readable; the operational layer makes you operable. Publish both, and keep both current.

swa:1:aHR0cHM6Ly9zaGFyZXdpdGhteWFnZW50LmFpL2Jsb2cvd3JpdGUteW91ci1maXJzdC1sbG1zLXR4dC5qc29u

Keep reading