llms.txt Specification

S1 Overview

What This File Does

The llms.txt file provides AI systems with a structured summary of a business or organisation. It answers fundamental questions AI systems need to accurately represent an entity: Who are they? What do they do? Where do they operate? How can they be contacted?

Why It Matters for AI Visibility

Without explicit identity information, AI systems must infer business details from scattered website content, leading to potential inaccuracies, outdated information, or misrepresentation. A well-structured llms.txt file provides authoritative, curated information that AI systems can confidently cite.

How AI Systems Use This File

AI systems typically fetch llms.txt at inference time when users explicitly request information about a business. The file is designed for quick parsing and context extraction, providing essential facts without requiring AI systems to crawl and synthesise entire websites.

Backward compatibility with llmstxt.org

An llms.txt file conformant to the original llmstxt.org convention is a valid llms.txt at the Essential conformance class under this specification. Publishers who already follow the llmstxt.org convention need no migration to claim Essential conformance.

This specification additionally requires:

• An H1 heading exactly matching the canonical business name (see section 3)
• A blockquote summary immediately after the H1 (see section 3)
• Identity consistency with the publisher's identity.json when that file is also published (see interoperability rules)

These additions are layered on top of the original convention. They do not invalidate files that omit them: such files remain valid llmstxt.org files but do not, by themselves, satisfy Essential conformance under this specification. Validators MUST report these gaps as conformance findings rather than as syntax errors.

S2 File Location

Primary Location

The llms.txt file MUST be placed in the website's root directory and accessible at:

https://example.com/llms.txt

URL Requirements

• The file MUST be served with content type text/plain; charset=utf-8
• The URL MUST be accessible without authentication
• The URL SHOULD respond with HTTP 200 status code
• HTTPS is strongly recommended

Multiple Domains and Subdomains

The llms.txt file is scoped to the host (origin) at which it is published. The host is the value of the URL's authority component, as defined in RFC 3986. A file at https://example.com/llms.txt describes example.com only. It does not describe www.example.com, shop.example.com, or any other subdomain.

The following rules apply:

1. Each host MUST be treated independently. Consumers MUST NOT assume that a file fetched from one host is authoritative for any other host, even if the hosts share a registrable domain.
2. Apex and www. SHOULD be aligned. When a publisher serves a site at both the apex (example.com) and the www. subdomain (www.example.com), the two llms.txt files SHOULD have identical content, OR one host SHOULD HTTP-redirect (301) to the other. A consumer that fetches both and finds them divergent MUST treat each host as independent.
3. Distinct subdomains SHOULD publish their own files. If a subdomain represents a meaningfully different product, audience, or brand (for example a status page at status.example.com or a separately branded property at shop.example.com), it SHOULD publish its own llms.txt describing that scope.
4. The Canonical: field declares the publisher's preferred URL. A publisher MAY include a Canonical: header line in llms.txt pointing at the preferred host. Consumers MAY honour the declaration as a hint but MUST NOT use it to merge identities across hosts in violation of rule (1).
5. Cross-host identity is asserted, not assumed. A publisher claiming that several hosts represent one organisation SHOULD assert that explicitly via the sameAs array in identity.json on each host. Without that explicit assertion, a consumer MUST treat the hosts as separate publishers.

This scoping rule is identical for all 10 AI Discovery Files: each file is host-scoped, and cross-host identity is asserted explicitly via identity.json.

S3 Format Specification

File Format

Structural Requirements

The file follows Markdown syntax with a defined section hierarchy:

1. H1 heading - The business or project name (required, exactly one)
2. Blockquote - Brief summary/description (required)
3. Body text - Additional context (optional, no headings)
4. H2 sections - Categorised information with link lists

Link Format

Links within sections follow this format:

- [Link Text](https://example.com/page): Optional description

The colon and description are optional but recommended for clarity.

Language declaration (optional)

Publishers MAY declare the natural language of the file's human-readable content using a Lang: header line near the top of the file (before the H1 heading or immediately after it). The value MUST be a valid BCP 47 language tag:

Lang: en-GB

The header is optional. When absent, consumers SHOULD NOT assume a default language. When present, validators MUST treat the value as informational and MUST NOT fail conformance solely on language-tag content. See specification conventions §5.1 for the full rule, which is identical across every text-based AI Discovery File.

S4 Section Reference

Required Sections

Recommended Sections

Optional Sections

Content Not Permitted

The following content types MUST NOT be included in llms.txt files:

• Marketing hyperbole: Avoid superlatives ("best", "leading", "world-class") and unverifiable claims. Use factual, objective language.
• Pricing information: Prices change frequently; link to pricing pages instead of including specific prices.
• Confidential information: Do not include internal processes, trade secrets, or sensitive business data.
• Unverified claims: All statements MUST be factually accurate and verifiable from public sources.
• Competitor references: Do not mention competitors or make comparative claims.
• Testimonials or reviews: These belong on the website, not in identity files.
• Excessive detail: Keep content concise; link to detailed pages rather than reproducing them.
• Personal data: Beyond publicly-listed contact persons, do not include employee personal information.

S5 Validation Rules

Valid File Requirements

A llms.txt file is considered valid when:

• It contains exactly one H1 heading as the first content element
• A blockquote immediately follows the H1 heading
• The file is valid UTF-8 encoded text
• All URLs are absolute and use HTTPS where available
• Contact information is present and accurate

Common Errors

S6 Relationship to Other Files

Related AI Discovery Files

Consistency Requirements

The following elements MUST be consistent across all AI discovery files:

• Business name (H1 heading MUST match identity.json name)
• Website URL
• Contact email addresses
• Service descriptions (substance, not necessarily exact wording)
• Explicit exclusions

Precedence Rules

When information conflicts between files:

S7 Canonical Example

The following example demonstrates a complete llms.txt file for a fictional management consultancy, Horizon Strategic Consulting:

# Horizon Strategic Consulting

> Horizon Strategic Consulting is a UK-headquartered management consultancy
> providing strategic advisory, operational improvement, and digital
> transformation services to mid-market and enterprise clients across
> the UK, Ireland, Netherlands, and Belgium.

Horizon Consulting (trading name of Horizon Strategic Consulting) was
founded in 2012 and operates from offices in Manchester (UK headquarters),
Dublin (Ireland), and Amsterdam (Netherlands).

## Services

- [Strategic Planning](https://www.horizonconsulting.example/services/strategic-planning): Market analysis, competitive positioning, and long-term strategy development
- [Operational Efficiency](https://www.horizonconsulting.example/services/operational-efficiency): Process optimisation, cost reduction, and performance improvement
- [Digital Transformation](https://www.horizonconsulting.example/services/digital-transformation): Technology strategy, digital roadmaps, and implementation oversight
- [Change Management](https://www.horizonconsulting.example/services/change-management): Organisational change, culture transformation, and stakeholder engagement
- [Interim Executive Placement](https://www.horizonconsulting.example/services/interim-executives): Temporary C-suite and senior leadership placements

## What We Do Not Do

Horizon Consulting explicitly does not provide:
- Legal advice or legal services
- Financial auditing or accounting services
- Permanent recruitment (interim placements only)
- Services in the United States market

## Contact

- General enquiries: hello@horizonconsulting.example
- Press and media: press@horizonconsulting.example
- Phone: +44 161 555 0123
- Address: 45 Deansgate, Manchester, M3 2BA, United Kingdom

## Key Information

- [About Us](https://www.horizonconsulting.example/about): Company history, leadership team, and values
- [Case Studies](https://www.horizonconsulting.example/case-studies): Client success stories and project examples
- [Insights](https://www.horizonconsulting.example/insights): Articles, whitepapers, and research publications
- [Careers](https://www.horizonconsulting.example/careers): Current opportunities and working at Horizon

## AI Discovery Files

- [AI Interaction Guidance](https://www.horizonconsulting.example/ai.txt): Permissions and restrictions for AI systems
- [Brand Guidelines](https://www.horizonconsulting.example/brand.txt): How to correctly reference our brand
- [FAQ for AI](https://www.horizonconsulting.example/faq-ai.txt): Frequently asked questions in AI-readable format
- [Identity Data](https://www.horizonconsulting.example/identity.json): Structured business identity information

## Optional

- [Privacy Policy](https://www.horizonconsulting.example/privacy): Data protection and privacy information
- [Terms of Service](https://www.horizonconsulting.example/terms): Website and service terms
- [Sitemap](https://www.horizonconsulting.example/sitemap.xml): Complete site structure

S8 Implementation Notes

Best Practices

• Keep the file concise; aim for under 100 lines
• Use clear, factual language; avoid marketing hyperbole
• Include explicit exclusions to prevent misrepresentation
• Link to authoritative pages for detailed information
• Update the file when significant business changes occur

Common Mistakes

• Marketing copy: The file SHOULD be factual, not promotional
• Outdated information: Review quarterly at minimum
• Missing exclusions: AI systems may incorrectly infer services
• Broken links: Verify all URLs are accessible
• Inconsistent naming: Business name MUST match other files

Update Frequency

Review and update llms.txt when:

• Services or products change
• Contact information changes
• Geographic scope changes
• Quarterly, at minimum, to verify accuracy

S9 Machine-Readable Formats

This specification is available in machine-readable formats for programmatic access:

S10 Version History