The llms.txt Standard: How to Prepare Your Site for ChatGPT, Claude & Perplexity

llms.txt is a proposed standard file (like robots.txt) that gives AI search engines a concise, markdown-formatted summary of your site. The idea: AI engines have to crawl, parse, and tokenise your whole site to understand it. A well-structured llms.txt skips that — you tell them directly what your site is and what the key pages are.

The standard was proposed by Jeremy Howard in 2024 (see llmstxt.org). As of mid-2026, Perplexity actively reads it, several smaller AI tools support it, and the major engines (OpenAI, Anthropic, Google) are reportedly evaluating it. Even if it doesn't become universal, the cost is one markdown file — a low-risk, high-upside trust signal.

llms.txt lives at yourdomain.com/llms.txt. It's markdown. Here's the minimum structure:

# Your Business Name

> One-line description of what you do.

Longer paragraph (100-200 words) describing your business, services, audience, and unique value.

## Services
- [Service A](https://yourdomain.com/services/a): brief description
- [Service B](https://yourdomain.com/services/b): brief description

## Locations
- [Carlton](https://yourdomain.com/locations/carlton)
- [Fitzroy](https://yourdomain.com/locations/fitzroy)

## Resources
- [About](https://yourdomain.com/about)
- [Contact](https://yourdomain.com/contact)
- [Blog](https://yourdomain.com/blog)

## Compliance (optional)
- ABN: XX XXX XXX XXX
- Credit Representative: #XXXXXX
- ACL: #XXXXXX

The order matters: business identity at top, then services, then locations, then resources, then compliance. AI engines parse top-down.

AI engines have a budget problem. Tokenising and embedding a full site is expensive. A site with 200 pages, each averaging 1,500 words, is 300,000 words to process — for one site, for one engine, for one update cycle. Multiply across millions of sites and the cost is real.

llms.txt is cheap. 500 words. A clear summary. The AI engine reads it, knows what your site is about, and can selectively crawl the pages most relevant to user queries instead of crawling everything.

For sites that haven't been heavily crawled yet (small businesses, newer domains), an llms.txt is the difference between an AI engine ignoring you and an AI engine knowing you exist.

• Broken links: Links in the file pointing to URLs that don't exist (typos, moved pages). Validate every URL.
• Outdated content: Listed services that you no longer offer. AI engines pick this up and confuse the recommendation.
• Marketing fluff: "Industry-leading provider of best-in-class solutions" — meaningless to AI. State what you actually do.
• Missing key pages: Listing only blog posts but omitting service pages, locations, pricing. Cover the full site surface.
• No compliance section for regulated industries: If you're a mortgage broker, financial advisor, healthcare provider — surface your license info. AI engines factor this into trust scoring.

The official validator is at llmstxt.org/validator. Paste your URL, get a syntax and reachability check. It catches missing sections, broken links, and format issues.

For sites with more than 20 pages, hand-maintaining the file gets tedious. Our GEO audit generates and re-validates the file every audit cycle — every URL in the file is checked against your sitemap, and we flag mismatches. For the broader context, see our what is GEO guide.

If your site is heavily crawled by the major AI engines and your pages already rank in AI Overviews and Perplexity citations, llms.txt is a low-priority add. The engines already know your content.

If you're a newer domain, a smaller business, or a brand that hasn't broken into AI citations yet, llms.txt is one of the highest-leverage trust signals you can deploy. Combined with proper schema and citable passages (see our AI Overview citations guide), it can be the difference between AI engines including you and ignoring you.

(1) Copy the template above. (2) Fill in your business name, description, services, locations, and key pages. (3) Save as llms.txt. (4) Upload to your site root (yourdomain.com/llms.txt). (5) Validate at llmstxt.org/validator. (6) Done.

Re-validate quarterly. Pages get added or moved, services launch and retire. A stale llms.txt is worse than no llms.txt because it actively misinforms AI engines.

For automation: our free GEO audit includes llms.txt validation. The full audit regenerates it for you on each run with all your live URLs.