Specifications

The files, defined

Eleven files across three layers. Each one has a single job, and each links to its full specification document in the project repository.

Specifications on GitHub Templates repository Copy the templates

Layer 1

Core Identity

Establishes who you are and makes it readable by both machines and people. Start here — everything else refers back to it.

llms.txt · llm.txt · llms-full.txt · llms.html · identity.json · ai.json

Layer 2

Guidance & Control

Sets the rules: when to recommend you, how to name you, and how AI crawlers should treat your content.

ai.txt · brand.txt · robots-ai.txt

Layer 3

Enhancement

Adds verified answers and technical context, so AI systems have something specific to quote instead of inferring.

faq-ai.txt · developer-ai.txt

What the priorities mean

Required: Publish this. Without it there is no authoritative identity to read.
Recommended: Publish this unless you have a specific reason not to.
Optional: Publish when you want the extra control it gives.
Conditional: Publish only when the condition applies to your site.

Layer 1

Core Identity

llms.txt

Required

Your identity, in the format AI systems read first.

The canonical AI-readable identity file and the single source of truth for this file set. It states who you are, what you offer, where you operate and how to reach you. Every other file must align with the Core Identity fields declared here.

Format: Markdown, UTF-8, served as text/plain
Location: https://example.com/llms.txt

A valid file has

An H1 with your brand name
A one-sentence factual summary as a blockquote
The Core Identity block, word for word consistent across files
A Last updated date

Also worth reading

llmstxt.org The upstream format specification by Jeremy Howard, which this file follows.
Lighthouse: llms.txt audit Chrome DevTools audits for this file under Agentic browsing. A 404 is reported as Not Applicable, since publishing it is currently optional — but a server error is flagged.

Full specification Template

llm.txt

Required

A compatibility copy for systems that ask for the singular name.

Some AI systems request /llm.txt by convention. This file makes sure those requests resolve rather than 404. It must point at the same source of truth as llms.txt so the two never send conflicting signals.

Format: 301 redirect to llms.txt, or an identical copy
Location: https://example.com/llm.txt

A valid file has

Same content as llms.txt, or a 301 to it
Never a second, diverging source of truth

Full specification Template

llms-full.txt

Conditional

The expanded companion, when one file is not enough.

Use when a concise llms.txt cannot cover your important public pages clearly.

Follows the same Markdown format as llms.txt but carries deeper context and broader link coverage. It exists so llms.txt can stay short and authoritative rather than growing into a sitemap.

Format: Markdown, UTF-8, served as text/plain
Location: https://example.com/llms-full.txt

A valid file has

The same Core Identity block as llms.txt
Broader link coverage, not contradictory detail
A link back to llms.txt as canonical

Full specification Template

llms.html

Recommended

The same information, readable by people.

A browser-friendly HTML version of your llms.txt content, so a human who finds the file has something legible to read. It is a presentation layer — it must not state anything llms.txt does not.

Format: HTML5, optionally with Schema.org structured data
Location: https://example.com/llms.html

A valid file has

Content that matches llms.txt
No tracking scripts or advertising
A link to the canonical llms.txt

Full specification Template

identity.json

Required

Your identity as structured, parseable data.

Schema.org-aligned identity facts in JSON: names, locations, contact points, registration details. It makes the same facts easy to parse and validate rather than extract from prose.

Format: Strict JSON, UTF-8, served as application/json
Location: https://example.com/identity.json

A valid file has

Core Identity fields matching llms.txt
Schema.org-aligned structure
Contact points
A lastUpdated value

Full specification Template

ai.json

Recommended

Structured identity plus recommendation signals.

Adds machine-readable recommendation and policy context alongside your identity data, complementing the plain-text files with explicit structured signals.

Format: Strict JSON, UTF-8, served as application/json
Location: https://example.com/ai.json

A valid file has

Core Identity fields matching llms.txt
Recommendation and policy signals
Values consistent with ai.txt

Full specification Template

Layer 2

Guidance & Control

ai.txt

Recommended

When AI should recommend you — and when it should not.

Plain-text eligibility and exclusion guidance. It tells AI systems which situations you are a good answer for, and which you are explicitly not, so you stop being suggested for work you do not do.

Format: Plain text with headed sections
Location: https://example.com/ai.txt

A valid file has

The Core Identity block
Recommend / do not recommend guidance
Clear, specific and actionable statements

Full specification Template

brand.txt

Recommended

How to name you correctly — and how not to.

Official name, accepted short forms, pronunciation, voice and tone, and the forms that must never be used. This is the file that stops an AI inventing an abbreviation or merging you with a similarly named company.

Format: Plain text with headed sections
Location: https://example.com/brand.txt

A valid file has

Official and legal names
Naming rules and forms to avoid
Voice and tone guidance

Full specification Template

robots-ai.txt

Optional

Citation preferences and crawler-level guidance for AI.

Communicates content citation preferences, freshness expectations and crawler-level hints for AI indexing. It supplements robots.txt and never overrides it.

Format: robots.txt-style syntax
Location: https://example.com/robots-ai.txt

A valid file has

Citation and freshness preferences
Rules consistent with robots.txt, which still wins

Full specification Template

Layer 3

Enhancement

faq-ai.txt

Recommended

Verified answers, written to be quoted.

Real customer questions with factual, verifiable answers. Unlike an FAQ page built for browsing, this is written on the assumption an AI will lift the answer word for word.

Format: Structured plain text
Location: https://example.com/faq-ai.txt

A valid file has

The Core Identity block
Questions drawn from real enquiries
Answers supported by public content

Full specification Template

developer-ai.txt

Conditional

Technical context for developer-facing questions.

Use when your product exposes APIs, SDKs or integration workflows.

API availability, authentication, rate limits and where the real documentation lives. It provides context and points onward rather than duplicating your API reference.

Format: Plain text with headed sections
Location: https://example.com/developer-ai.txt

A valid file has

The Core Identity block
API status and authentication summary
Links to real documentation
No credentials, internal URLs or unreleased features

Full specification Template

Consistency

How the files relate

The value of this file set comes from the files agreeing with each other. Four rules keep them aligned.

llms.txt is the source of truth

Every file that carries the Core Identity block must repeat it word for word. Where a fact differs between files, llms.txt is the one that is correct and the others need fixing.

llm.txt must never diverge

It is a redirect or an identical copy, never a second identity. Two files that drift apart send conflicting signals, which is worse than publishing neither.

robots.txt still wins

robots-ai.txt adds AI-specific citation and freshness preferences on top of your existing robots.txt. It supplements, it does not override.

Omit, or answer explicitly

Leave out optional fields that do not apply. But where a reader might wrongly assume otherwise, say so outright — "No public API" beats an ambiguous blank.

Files carrying the Core Identity block

These must repeat the same identity lines, character for character. A mismatch here is the single most common validation failure.

llms.txt · llms-full.txt · identity.json · ai.json · ai.txt · faq-ai.txt · developer-ai.txt

Summaries on this page are written for readability. The specification documents in the project repository are the authority — where the two differ, the repository is correct. llms.txt follows the format published at llmstxt.org .