The llms.txt standard — declare your authoritative content to AI

What is llms.txt?

llms.txt is a plain-Markdown file at the root of your domain (/llms.txt) that lists your most authoritative pages, organized by section. It tells AI crawlers and the language models behind ChatGPT, Claude, and Perplexity which content on your site is the canonical source.

Think of it as robots.txt for meaning rather than access. robots.txt says "crawl this, don't crawl that." llms.txt says "this is the canonical version of our content — quote this." The standard was proposed by Jeremy Howard (Answer.ai) in September 2024.

There is a companion file: llms-full.txt. The short version (llms.txt) lists URLs with descriptions; the full version (llms-full.txt) concatenates the actual page contents for one-shot model ingestion. Either or both belong at the domain root.

What does a real llms.txt look like?

Minimal example for a fictional Acme Widgets storefront. The H1 names the site; the blockquote summarizes it; the H2 sections cluster the URLs; each URL has a short description after the colon.

# Acme Widgets

> The leading direct-to-consumer widget marketplace. 12,000 SKUs across 8 categories.

## Main

- [Home](https://acme.example/): Catalog entry point
- [About](https://acme.example/about): Company overview, leadership, contact
- [Shipping policy](https://acme.example/shipping): Rates, regions, transit times

## Products

- [Widgets — All](https://acme.example/widgets): Filterable widget catalog
- [Best sellers](https://acme.example/best-sellers): Top-30 by 30-day sales

## Support

- [Returns](https://acme.example/returns): 30-day policy
- [FAQ](https://acme.example/faq): 40 most common buyer questions

## Optional

- [Press](https://acme.example/press): Press kit + media contact
- [Investors](https://acme.example/investors): Quarterly reports

For a live example built from real production data, visit /llms.txt on this very domain — it regenerates every hour from our database and lists current public audits, tools, learn surface, and community.

How do I write a good llms.txt?

1. 1. Open with H1 + blockquote summary

   First line is the H1 — your site name. Second line is a > blockquote that summarizes who you are and what the site delivers in one sentence.

2. 2. Cluster by content type with H2 sections

   Use H2 headings for major content classes — typically "Main", "Products" or "Services", "Docs", "Support", "Optional". Optional is for nice-to-have URLs the model can skip.

3. 3. List 3–10 Markdown links per section

   Use standard Markdown link syntax: - [Page name](https://example.com/page): one-line description after the colon. Description is what the model reads when deciding to follow the link.

4. 4. Order by authority, not alphabetically

   The model reads top to bottom and weighs the order. Put your strongest evergreen pages first. Long-tail or niche pages can live in Optional.

5. 5. Keep it under 10 KB

   Curate ruthlessly. A 10,000-URL llms.txt is identical to no llms.txt — the model cannot extract authority signal from noise. Top 20–50 URLs is the right shape.

6. 6. Serve at /llms.txt with text/markdown

   Deploy at https://yourdomain.com/llms.txt with Content-Type: text/markdown; charset=utf-8. Cache for an hour or so (we use Cache-Control: public, max-age=3600, stale-while-revalidate=86400 on our own).

Which AI engines respect llms.txt today?

Why is llms.txt worth your 30 minutes?

• Less than 1% of indexed domains have llms.txt in 2026 — early-adopter window
• Anthropic and Perplexity treat it as a primary signal — measurable citation lift
• No SEO downside — Google ignores the file, no ranking impact
• 30-minute investment for a 2–3 year asymmetric upside
• Signals expertise — implementers notice when your AEO posture is current
• Composability — pairs with structured data, llms-full.txt, direct-answer format

llms.txt frequently asked

• What is llms.txt?

  llms.txt is a plain-Markdown file at the root of your domain (/llms.txt) that lists your most authoritative pages, organized by section. It tells AI crawlers and the language models behind ChatGPT, Claude, and Perplexity which content on your site is canonical — robots.txt for meaning instead of access. The standard was proposed by Jeremy Howard (Answer.ai) in September 2024.

• Is llms.txt mandatory? Will I lose AI traffic without it?

  Not mandatory yet — none of the major AI engines treat it as required. But adopting it puts you in a tiny minority (less than 1% of indexed domains in 2026) that AI crawlers can disambiguate efficiently. The cost is ~30 minutes of writing; the upside is measurably higher citation share for the next 2–3 years until the standard becomes table stakes.

• What is the difference between llms.txt and llms-full.txt?

  llms.txt is the INDEX — a short Markdown file (typically 1–4 KB) that lists your top URLs by section. llms-full.txt is the EXPANDED file (often 100 KB to several MB) that concatenates the actual full text of those URLs for one-shot model ingestion. Many sites only ship llms.txt and let the model fetch each linked URL on demand. Both files belong at the domain root.

• Which AI engines actually respect llms.txt today?

  As of mid-2026 the most explicit adopters are Anthropic (Claude) and Perplexity. OpenAI's GPTBot honors the URL list as a hint for crawl prioritization but does not commit to following it strictly. Google has not formally adopted the standard for AI Overviews. Even partial adoption is enough to move the needle — the cost to ship is so low that the asymmetric upside is worth it.

• How do I write a good llms.txt file?

  Start with an H1 (your site name) and a one-line blockquote summary. Add H2 sections for major content types ("Main", "Products", "Docs", "Support", "Optional"). Under each, list 3–10 links in Markdown format with a short description after the colon. Order by importance — the model reads top to bottom. Keep the whole file under 10 KB. See our free generator for a guided builder.

• Should llms.txt be the same as my sitemap.xml?

  No. sitemap.xml is for traditional search engines and is exhaustive (every indexable URL). llms.txt is for AI engines and is curated (only your top 20–50 authoritative URLs). A 10,000-URL llms.txt defeats the purpose — the model can't distinguish authoritative content from incidental content. Curate.

• Will llms.txt affect my Google ranking?

  Not directly. Google's crawler ignores llms.txt; the file influences AI engines that operate independently of Google rankings. There is no SEO penalty for adopting llms.txt and no ranking lift from doing so. The benefit is strictly downstream in the AI-citation graph (ChatGPT, Claude, Perplexity), not in Google's blue-link SERP.

• Does SeoChatAI have an llms.txt I can look at?

  Yes — we eat our own dogfood. Visit /llms.txt on this domain to see a live DB-driven example that lists our main surfaces, free tools, learn hub, community, public APIs, and recently audited example domains. It regenerates every hour from real production data.

Keep going