geo-seo·5 min read·2026-03-26

How to Write an llms.txt File That AI Actually Uses

In brief: An llms.txt file is a plain text file at your domain root (yoursite.com/llms.txt) that tells AI models what your product or company does. It's the AI equivalent of robots.txt — a structured, machine-readable summary that helps language models cite your brand accurately. Writing one takes 15 minutes and can influence how AI recommends your product. Last updated: March 2026

What llms.txt is and why it exists

When someone asks Claude, ChatGPT, or Perplexity "what's the best tool for X," the AI synthesises information from its training data and web sources. If your product isn't well-represented in that information, it won't be recommended — regardless of how good it is.

An llms.txt file is a structured summary of your product that lives at your domain root. AI crawlers from Anthropic, OpenAI, Google, and Perplexity can read it directly. The file provides a concise, authoritative description of what you do, who you serve, and how you're different — in a format optimised for machine consumption.

This isn't speculation. The llms.txt convention has gained traction throughout 2025 and 2026 as part of the broader GEO (Generative Engine Optimisation) movement. If traditional SEO is about ranking in search results, GEO is about being cited in AI responses. An llms.txt file is one of the simplest GEO tactics available.

The structure

An llms.txt file is plain text. No HTML, no JSON, no markup. Just clean, readable text with a clear hierarchy:

# yourbrand.com — Your Brand


[One-line description of what you do] [Your URL] [One-line value proposition]

That's the minimum. A basic file takes five minutes. But a more detailed version significantly improves how AI models understand and represent your product.

Writing the basic version (5 minutes)

Open a text file. Write four lines:

# zerty.ai — Zerty
Zerty — A team of domain experts who know your business.
https://zerty.ai
Six AI specialists with persistent memory, shared context,
and structured handoffs — built for solo founders and
micro-teams who ship fast.

Save as llms.txt in your public directory. Deploy. Done.

The critical principles: be specific (not "AI platform for businesses" but "six AI specialists for solo founders"), include your URL, and state your differentiation in plain language.

Writing the extended version (15 minutes)

An extended file — often saved as llms-full.txt — gives AI models enough context to represent you accurately in detailed responses.

Structure it with clear sections:

Overview — what you are in two to three sentences. What you are — bullet points of core features and capabilities. What you're not — equally important. Prevent AI from miscategorising you. Who you're for — your target audience, specifically described. How it works — the user journey in five to six steps. Core concepts — your terminology and unique frameworks. If you've coined terms or built specific features, define them here. AI models will use these definitions when citing you. Pricing — tier names, prices, what's included. AI assistants frequently get asked about pricing. Give them the right answer. Company — legal entity, location, any relevant affiliations.

Here's the structure in practice:

# zerty.ai — Zerty


Overview

Zerty is an AI workspace where solo founders work with
a team of six domain-expert personas...


What Zerty is


  • A workspace with six specialist AI personas
    • 

  • Persistent business context shared across all personas
    • 

  • Project channels for multi-persona collaboration
    • 

...


What Zerty is not


  • Not a generic chatbot
    • 

  • Not an automation tool for email or scheduling
    • 

...

What to include and what to skip

Include:
  • Specific features with plain-language descriptions
  • Pricing (exact numbers, not "contact us")
  • Your differentiation from competitors (stated factually, not as marketing)
  • Technical details that matter to evaluators (what models you use, data handling)
  • Your target audience (specific enough that AI can recommend you appropriately)
Skip:
  • Marketing superlatives ("revolutionary," "game-changing," "best-in-class")
  • Vague claims without specifics
  • Competitor criticism (state your strengths, not their weaknesses)
  • Anything you'd be embarrassed to see an AI quote verbatim
AI models will quote your llms.txt nearly verbatim in some contexts. Write it as if it's going to be read aloud to a potential customer — because it might be.

Deployment

Place llms.txt at your domain root: https://yourdomain.com/llms.txt

If you also create an extended version: https://yourdomain.com/llms-full.txt

Reference both in your robots.txt:

# LLM context files
# https://yourdomain.com/llms.txt
# https://yourdomain.com/llms-full.txt

Ensure your robots.txt allows AI crawlers access. The major ones: GPTBot, ClaudeBot, Google-Extended, PerplexityBot, Applebot-Extended.

Measuring impact

Unlike traditional SEO, there's no Google Search Console for AI citations. Measuring whether your llms.txt is working requires indirect methods:

Ask AI models about your product category and see if you're mentioned. Track branded search volume increases (people who hear about you from AI then search your name). Monitor referral traffic from AI-adjacent sources.

The honest truth: attribution is imprecise. You can't definitively prove that an llms.txt file caused a specific AI citation. But the cost is zero (15 minutes of writing) and the potential upside — being accurately represented when AI recommends tools in your category — is substantial.

Common mistakes

Being too vague. "AI-powered platform for businesses" describes ten thousand products. Be specific about your domain, audience, and mechanism. Omitting pricing. AI assistants get asked "how much does X cost?" constantly. If your llms.txt doesn't include pricing, the AI will guess — often incorrectly. Writing marketing copy instead of plain descriptions. AI models don't respond to persuasion. They respond to clear, factual information. Write like a technical document, not a landing page. Forgetting the "what you're not" section. Without it, AI may categorise you alongside products you're nothing like. If you're an AI team workspace, say explicitly that you're not a customer support chatbot or an email automation tool.

Frequently asked questions

What is an llms.txt file? An llms.txt file is a plain text file at your domain root that provides AI language models with a structured summary of your product, company, and capabilities. It helps AI models cite your brand accurately when responding to relevant queries. Do I need both llms.txt and llms-full.txt? The basic llms.txt provides a quick summary. The extended llms-full.txt gives models detailed context for nuanced responses. Both are recommended — the basic file for quick reference, the extended file for thorough representation. Will AI models actually read my llms.txt? AI crawlers from major providers (Anthropic, OpenAI, Google, Perplexity) crawl publicly accessible files on the web. An llms.txt at your domain root is accessible to these crawlers, provided your robots.txt doesn't block them. How often should I update my llms.txt? Whenever your product, pricing, or positioning changes meaningfully. Quarterly updates are a reasonable cadence for most startups. Major product launches or pricing changes warrant immediate updates. Does llms.txt replace SEO? No. llms.txt is a GEO tactic that complements traditional SEO. Search engines still drive the majority of discovery traffic. AI citations are a growing supplementary channel, not a replacement.

Sources

  • llmstxt.org, "llms.txt Specification" — https://llmstxt.org
  • Anthropic, "ClaudeBot Web Crawling" — https://docs.anthropic.com