llms.txt Implementation Step-by-Step

What llms.txt is and why it matters

llms.txt is a 2025 community standard for a single text file (served at /llms.txt) that points AI assistants to the canonical, machine-readable summary of your site. Think robots.txt for AI assistants, but with positive guidance rather than just allow/disallow. By mid-2026 the major AI assistants (ChatGPT, Claude, Perplexity, and partially Gemini) use llms.txt as a freshness and authority signal.

Step 1: Decide What to Include

A llms.txt is a directory, not a dump. Include the high-signal pages an AI assistant should know about:

• Your homepage and core product pages.
• Your most-canonical reference pages (pricing, docs, comparison hubs).
• Your top-cited research or data pages.
• Your llms.txt-specific page summarising your business in 1-3 paragraphs.

Avoid linking to every blog post. AI assistants will discover those through the underlying linked pages. Use llms.txt to elevate the spine of your site, not duplicate the sitemap.

Step 2: Use the Standard Structure

# [Brand Name]

> [One-sentence elevator description with named entities and a specific number]

## About

[2-3 sentence paragraph: who you are, what you do, when founded, where headquartered. Plain prose, no marketing fluff.]

## Products

- [Product 1](https://example.com/product1): One-line description with a specific differentiator.
- [Product 2](https://example.com/product2): One-line description.

## Documentation

- [API reference](https://example.com/docs/api): Brief description.
- [Pricing](https://example.com/pricing): Brief description.

## Research / Data

- [Headline research page](https://example.com/research/x): Brief description.
- [Headline data page](https://example.com/research/y): Brief description.

## Policies

- [Privacy](https://example.com/privacy)
- [Terms](https://example.com/terms)
- [AI use policy](https://example.com/ai-policy)

## Contact

- Support: support@example.com
- Press: press@example.com

Step 3: Host at /llms.txt

Serve the file at the root: https://yourbrand.example/llms.txt. Use text/plain; charset=utf-8 Content-Type. Return 200 OK; do not return 301/302 redirects to a fancy URL. Many AI assistants only follow the canonical /llms.txt path.

Step 4: Add an Extended Version at /llms-full.txt (Optional)

The optional /llms-full.txt file mirrors /llms.txt but inlines key page content (not just URLs) so AI assistants can extract directly without an additional crawl. Useful for sites with sensitive crawl budget or pages that AI bots have trouble parsing. Keep llms-full.txt under ~500KB.

Step 5: Worked Example — SaaS

# Acme Analytics

> Acme Analytics is a cloud analytics platform serving 12,000+ customers including 38 Fortune 500 companies, founded 2018, headquartered in San Francisco.

## About

Acme Analytics builds a unified analytics platform combining data warehousing, BI, and AI-powered insights. The company was founded in 2018 by Sarah Chen and Marcus Lee. Headquartered in San Francisco with offices in London and Tokyo.

## Products

- [Acme Cloud](https://acme.example/cloud): Cloud-native data warehouse with 99.99% SLA.
- [Acme Insights](https://acme.example/insights): AI-powered BI for non-technical users.
- [Acme Pipelines](https://acme.example/pipelines): No-code ETL with 200+ connectors.

## Documentation

- [API reference](https://acme.example/docs/api)
- [SDKs](https://acme.example/docs/sdks)
- [Quickstart](https://acme.example/docs/quickstart)

## Pricing

- [Pricing tiers](https://acme.example/pricing): Starter $99/mo, Pro $499/mo, Enterprise quote.

## Customers

- [Customer stories](https://acme.example/customers)
- [Acme vs Snowflake](https://acme.example/compare/snowflake)

## Policies

- [Privacy](https://acme.example/privacy)
- [Terms](https://acme.example/terms)
- [Security](https://acme.example/security)

Step 6: Worked Example — Ecommerce

# Acme Audio

> Acme is a consumer audio brand founded in 2014, headquartered in Brooklyn NY, with 18 retail stores in the US and a flagship online catalogue of 240+ products.

## About

Acme designs and sells consumer audio: wireless headphones, earbuds, portable speakers, and home audio gear. Founded in 2014 by Maria Ortiz. 18 retail stores plus an online catalogue serving 47 countries.

## Top Products

- [Acme X5 Headphones](https://acme.example/x5): $299. Active noise-cancelling wireless headphones with 40hr battery.
- [Acme Mini 2 Earbuds](https://acme.example/mini2): $149. True-wireless earbuds with adaptive EQ.

## Catalogue

- [All headphones](https://acme.example/c/headphones)
- [All earbuds](https://acme.example/c/earbuds)
- [Speakers](https://acme.example/c/speakers)

## Support

- [Returns](https://acme.example/returns): 30-day return policy, free shipping.
- [Warranty](https://acme.example/warranty): 2-year manufacturer warranty.
- [Contact](https://acme.example/contact)

Step 7: Validate

1. Confirm https://yourbrand.example/llms.txt returns 200 OK with Content-Type text/plain.
2. Open the file in a plain-text editor; confirm no HTML or JavaScript leaked in.
3. Check every URL in the file resolves with 200 OK.
4. Run the file through community validators (e.g., llms-txt.org if available).
5. Test a prompt in Claude or Perplexity asking about your brand; check if facts from the llms.txt show up correctly.

Step 8: Keep It Fresh

Update llms.txt whenever you launch a major product, change pricing, change leadership, or move HQ. Most teams review llms.txt monthly. Stale llms.txt is worse than no llms.txt because AI assistants trust the signal and propagate the stale information.

Common Mistakes

1. Treating llms.txt as a sitemap. It is a curated directory, not a comprehensive list.
2. Marketing copy in the about paragraph. Write plain facts with named entities and specific numbers.
3. Linking to login-gated pages. AI assistants cannot crawl behind auth.
4. No machine-readable updates. Pair llms.txt with Schema.org markup on the linked pages.
5. One-and-done. Brands that publish llms.txt once and forget it lose the signal advantage over time.