LLM-LD 1.0 Specification Fresh

Status: Draft | Published: February 2026 | License: CC BY 4.0 | Maintained by: CAPXEL

Abstract

LLM-LD (Large Language Model Linked Data) is a specification for making websites accessible, understandable, and actionable by artificial intelligence systems. It extends Schema.org vocabulary with AI-specific constructs that enable crawlers, agents, and AI-powered applications to efficiently discover, ingest, and interact with web content.

This specification defines:

A file format (llm-index.json) that serves as a complete machine-readable representation of a website
A vocabulary of properties optimized for AI consumption
Conformance levels that allow incremental adoption
A verification mechanism through the LLM Disco Directory

Introduction
Terminology
Conformance Levels
File Structure
Core Properties
Entity Properties
Agent Intelligence Properties
Verification Properties
Meta Properties
File Placement and Discovery
Relationship to Schema.org
Security Considerations
IANA Considerations
Examples
JSON Schema
Changelog

1. Introduction

1.1 Background

The proliferation of AI systems that consume web content -- including search engines, chatbots, autonomous agents, and retrieval-augmented generation (RAG) systems -- has created a need for websites to communicate with these systems in structured, predictable ways.

Traditional web standards (HTML, robots.txt, sitemap.xml) were designed for human readers and traditional search crawlers. While Schema.org provides structured data vocabulary, it focuses on describing entities rather than enabling AI interaction.

LLM-LD addresses this gap by providing:

Consolidation: A single file containing all information an AI needs
Actionability: Clear definitions of what actions are available
Context: Guidance for AI decision-making
Verification: Trust signals through directory membership

1.2 Goals

Enable AI systems to understand a website's complete offering in a single request
Provide clear, unambiguous action endpoints for agentic AI
Establish trust through verifiable directory membership
Maintain backwards compatibility with Schema.org
Support incremental adoption through conformance levels

1.3 Audience

Web developers implementing LLM-LD on websites
AI system developers consuming LLM-LD data
Tool developers building LLM-LD generators and validators
Directory operators integrating with the LLM Disco network

1.4 Document Conventions

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are interpreted as described in RFC 2119.

2. Terminology

AI Crawler -- An automated system that retrieves and indexes web content for AI applications. Examples include GPTBot, ClaudeBot, and PerplexityBot.

Agent -- An AI system capable of taking actions on behalf of a user, including navigation, form submission, and API calls.

Conformance Level -- A defined set of requirements that an implementation must satisfy. LLM-LD defines three levels: Crawl-Ready, Ingest-Ready, and Agent-Ready.

Entity -- A distinct thing described by the website, such as a product, service, person, or organization.

LLM Disco Directory -- A network of LLM-LD compliant websites maintained by CAPXEL, providing verification and discovery services.

Primary Entity -- The main entity that a website represents, typically an Organization or LocalBusiness.

Verification -- The process of confirming that a website is legitimately associated with its claimed identity through directory membership.

3. Conformance Levels

LLM-LD defines three conformance levels, each building upon the previous.

3.1 Level 1: Crawl-Ready

Purpose: Enable AI crawlers to discover and index website content.

Requirements:

MUST serve valid robots.txt allowing AI crawlers
MUST provide sitemap.xml listing all pages
SHOULD include Schema.org JSON-LD on individual pages

Files:

website.com/
├── robots.txt
├── sitemap.xml
└── [pages with Schema.org markup]

3.2 Level 2: Ingest-Ready

Purpose: Enable AI systems to understand the complete website from a single file.

Requirements:

MUST satisfy Level 1 requirements
MUST serve llm-index.json at the designated location
MUST include all Core Properties (Section 5)
SHOULD include applicable Entity Properties (Section 6)

Files:

website.com/
├── robots.txt
├── sitemap.xml
├── .well-known/llm-index.json
└── [pages with Schema.org markup]

3.3 Level 3: Agent-Ready

Purpose: Enable AI agents to take actions and make informed decisions.

Requirements:

MUST satisfy Level 2 requirements
MUST include llmld:actions with at least one actionable endpoint
SHOULD include Agent Intelligence Properties (Section 7)
SHOULD register with the LLM Disco Directory for verification

3.4 Conformance Declaration

Implementations SHOULD declare their conformance level in the llmld:conformance property:

json

{
  "llmld:conformance": {
    "level": 3,
    "level_name": "Agent-Ready",
    "validated": true,
    "validated_by": "https://llmdisco.com",
    "validated_at": "2026-02-02T00:00:00Z"
  }
}

4. File Structure

4.1 File Format

LLM-LD documents MUST be valid JSON conforming to RFC 8259.

The file MUST use UTF-8 encoding without BOM.

The recommended filename is llm-index.json.

4.2 Context Declaration

Every LLM-LD document MUST include a JSON-LD context declaration:

json

{
  "@context": [
    "https://schema.org",
    "https://llmld.org/v1"
  ],
  "@type": "llmld:AIWebsite",
  "@id": "https://example.com/llm-index.json"
}

The @context array:

MUST include "https://schema.org" for Schema.org compatibility
MUST include "https://llmld.org/v1" for LLM-LD vocabulary

The @type:

MUST be "llmld:AIWebsite"

The @id:

MUST be the canonical URL of the llm-index.json file

4.3 Property Namespacing

Properties defined by this specification use the llmld: prefix:

llmld:summary -- LLM-LD property
name -- Schema.org property (no prefix needed in context)

4.4 Size Considerations

Files under 100KB are optimal for most AI systems
Files over 500KB SHOULD provide a llmld:summary sufficient for basic understanding
Very large sites MAY split content across multiple files referenced from the main index

5. Core Properties

Core properties are REQUIRED for Level 2 and Level 3 conformance.

5.1 llmld:meta

Description: Metadata about the LLM-LD document itself.

Property	Type	Required	Description
`version`	String	Yes	LLM-LD spec version (e.g., "1.0")
`generated`	DateTime	Yes	ISO 8601 timestamp of generation
`generator`	String	No	Software that generated this file
`expires`	DateTime	No	When this data should be considered stale
`refresh_interval`	String	No	Suggested refresh frequency (e.g., "daily")
`language`	String	No	Primary language (BCP 47 code, e.g., "en-US")

json

{
  "llmld:meta": {
    "version": "1.0",
    "generated": "2026-02-02T12:00:00Z",
    "generator": "aso-generator/2.0",
    "expires": "2026-02-09T12:00:00Z",
    "refresh_interval": "weekly",
    "language": "en-US"
  }
}

5.2 llmld:site

Description: Core identity information about the website and organization.

Property	Type	Required	Description
`name`	String	Yes	Organization or website name
`legal_name`	String	No	Registered legal name
`type`	String	Yes	Primary type (e.g., "Business", "NonProfit")
`industry`	Array[String]	No	Industry categories
`description`	String	Yes	Brief description (1-2 sentences)
`tagline`	String	No	Marketing tagline or slogan
`founded`	String	No	Year founded
`domains`	Object	Yes	Associated domains

domains Object:

Property	Type	Required	Description
`primary`	URL	Yes	Main website URL
`ai`	URL	No	AI-specific subdomain if different
`app`	URL	No	Application URL
`docs`	URL	No	Documentation URL
`api`	URL	No	API endpoint URL

5.3 llmld:primaryEntity

Description: Reference to the main entity this website represents.

Property	Type	Required	Description
`@id`	String	Yes	Local reference ID (e.g., "#organization")
`@type`	String	Yes	Schema.org type
`name`	String	Yes	Entity name
`description`	String	No	Brief description
`url`	URL	No	Canonical URL

5.4 llmld:summary

Description: Condensed information for quick AI understanding.

Property	Type	Required	Description
`one_liner`	String	Yes	Single sentence description
`paragraph`	String	Yes	2-4 sentence overview
`key_facts`	Array[String]	Yes	3-10 bullet-point facts
`differentiators`	Array[String]	No	What makes this entity unique
`target_customers`	Array[String]	No	Ideal customer descriptions

json

{
  "llmld:summary": {
    "one_liner": "Example Company builds AI-powered marketing tools for businesses.",
    "paragraph": "Example Company is a marketing technology company based in Miami, Florida. Founded in 2020, we build tools that help businesses understand their markets and customers.",
    "key_facts": [
      "Founded in 2020",
      "2,500+ customers",
      "3 SaaS products",
      "SOC 2 Type II certified"
    ],
    "differentiators": [
      "Daily data updates (competitors update weekly)",
      "Privacy-first analytics"
    ],
    "target_customers": [
      "Small to medium businesses",
      "Marketing teams"
    ]
  }
}

5.5 llmld:pages

Description: Complete sitemap in structured format.

Page Object Properties:

Property	Type	Required	Description
`path`	String	Yes	URL path (e.g., "/about")
`title`	String	Yes	Page title
`type`	String	Yes	Page type (see Page Types below)
`url`	URL	Yes	Full URL
`schemas`	Array[String]	No	Schema.org types used on this page
`entity`	String	No	Reference to entity defined in this document
`description`	String	No	Brief page description
`lastmod`	Date	No	Last modification date

Page Types: homepage, about, product, service, listing, article, person, contact, pricing, faq, legal, other

5.6 llmld:actions

Description: Available actions that users or agents can take.

Structure:

json

{
  "llmld:actions": {
    "primary": [...],
    "purchase": [...],
    "contact": [...],
    "resources": [...]
  }
}

Action Object Properties:

Property	Type	Required	Description
`id`	String	Yes	Unique action identifier
`name`	String	Yes	Human-readable action name
`description`	String	No	What this action does
`url`	URL	Yes*	Action URL (*or `channels` for contact)
`type`	String	Yes	Action type
`priority`	Integer	No	Display priority (1 = highest)
`requires_auth`	Boolean	No	Whether authentication is required
`best_for`	Array[String]	No	Use cases this action suits

Action Types: signup, purchase, schedule, form, contact, navigate, download, api

5.7 llmld:contact

Description: Contact information organized by purpose.

Property	Type	Required	Description
`general`	ContactPoint	No	General inquiries
`sales`	ContactPoint	No	Sales team
`support`	ContactPoint	No	Customer support
`address`	PostalAddress	No	Physical address

ContactPoint Properties:

Property	Type	Required	Description
`email`	Email	No	Email address
`phone`	String	No	Phone number (E.164 format preferred)
`chat`	URL	No	Live chat URL
`calendar`	URL	No	Scheduling link
`hours`	String	No	Availability hours
`response_time`	String	No	Expected response time

6. Entity Properties

Entity properties describe the things a website offers.

6.1 llmld:products

Describes software applications, physical products, digital goods.

Product Properties:

Property	Type	Required	Description
`@id`	String	Yes	Unique identifier
`@type`	String	Yes	Schema.org type (Product, SoftwareApplication, etc.)
`name`	String	Yes	Product name
`description`	String	Yes	Product description
`pricing.model`	String	Yes	"subscription", "one-time", "usage", "free"
`pricing.plans`	Array[Plan]	No	Available plans
`trial.available`	Boolean	Yes	Whether trial is available
`trial.duration_days`	Integer	No	Trial length in days

6.2 llmld:services

Describes professional services, consulting, recurring services.

Property	Type	Required	Description
`@id`	String	Yes	Unique identifier
`@type`	String	Yes	Schema.org type (Service, etc.)
`name`	String	Yes	Service name
`description`	String	Yes	Service description
`service_type`	String	No	"one-time", "recurring", "retainer", "project"
`deliverables`	Array[String]	No	List of deliverables
`delivery_method`	String	No	"in-person", "remote", "hybrid"

6.3 llmld:people

Describes team members, founders, subject matter experts.

Property	Type	Required	Description
`@id`	String	Yes	Unique identifier
`@type`	String	Yes	"Person"
`name`	String	Yes	Full name
`title`	String	No	Job title
`description`	String	No	Bio or description
`expertise`	Array[String]	No	Areas of expertise
`credentials`	Array[String]	No	Certifications, degrees

6.4 llmld:locations

Describes physical locations.

Property	Type	Required	Description
`@id`	String	Yes	Unique identifier
`@type`	String	Yes	Schema.org type (LocalBusiness, etc.)
`name`	String	Yes	Location name
`address`	PostalAddress	Yes	Full address
`hours`	Object	No	Business hours by day
`phone`	String	No	Location phone number

6.5 llmld:content

Describes blog posts, articles, case studies.

Property	Type	Required	Description
`@id`	String	Yes	Unique identifier
`@type`	String	Yes	Schema.org type (Article, BlogPosting, etc.)
`headline`	String	Yes	Article headline
`description`	String	No	Article summary
`datePublished`	Date	No	Publication date
`author`	String	No	Reference to person entity
`topics`	Array[String]	No	Topic categories

7. Agent Intelligence Properties

Agent intelligence properties help AI agents make better decisions. All properties in this section are OPTIONAL.

7.1 llmld:context

Provides background information that helps AI understand decisions.

json

{
  "llmld:context": {
    "elevator_pitch": "We help businesses understand their competitive landscape using AI.",
    "origin_story": "Founded by marketers frustrated with manual competitive research.",
    "positioning": "Premium competitive intelligence for data-driven teams.",
    "ideal_customer": "Marketing teams at B2B SaaS companies with $10M+ in revenue."
  }
}

7.2 llmld:faq

Frequently asked questions with structured answers.

json

{
  "llmld:faq": [
    {
      "@type": "Question",
      "name": "How long does implementation take?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Basic implementation (Layer 1 + AI Discovery Page) takes 15-20 minutes."
      }
    }
  ]
}

7.3 llmld:trust

Trust signals and social proof.

json

{
  "llmld:trust": {
    "reviews": {
      "google": { "score": 4.8, "count": 127 },
      "g2": { "score": 4.7, "count": 89 }
    },
    "certifications": ["SOC 2 Type II", "GDPR Compliant"],
    "awards": ["Best Local SEO Tool 2025"]
  }
}

8. Verification Properties

8.1 llmld:verification

Records verification status from the LLM Disco Directory.

json

{
  "llmld:verification": {
    "status": "verified",
    "verified_by": "https://llmdisco.com",
    "verified_at": "2026-02-01T00:00:00Z",
    "next_verification": "2026-05-01T00:00:00Z",
    "conformance": {
      "level": 3,
      "level_name": "Agent-Ready"
    }
  }
}

9. Meta Properties

9.1 llmld:pointers

Links to related AI-layer files.

json

{
  "llmld:pointers": {
    "entities": "https://ai.example.com/entities.json",
    "knowledge_graph": "https://ai.example.com/knowledge.graph.json",
    "llms_txt": "https://ai.example.com/llms.txt",
    "ai_discovery": "https://www.example.com/ai-discovery"
  }
}

10. File Placement and Discovery

10.1 Well-Known URI (Recommended)

/.well-known/llm-index.json

The /.well-known/ path is the recommended location.

10.2 Root Path (Alternative)

/llm-index.json

10.3 Discovery via robots.txt

Implementations SHOULD advertise the file location in robots.txt:

# AI Index
LLM-Index: https://ai.example.com/.well-known/llm-index.json

10.4 Discovery via HTML Link Element

html

<link rel="llm-index"
      type="application/json"
      href="https://ai.example.com/.well-known/llm-index.json"
      title="LLM-LD Index" />

10.5 HTTP Headers

Servers MAY include the location in HTTP response headers:

X-LLM-Index: https://ai.example.com/.well-known/llm-index.json

10.6 robots.txt Declaration

The robots.txt file SHOULD reference the llm-index.json location.

11. Relationship to Schema.org

LLM-LD extends Schema.org rather than replacing it. The llmld: namespace adds AI-specific properties that Schema.org does not provide:

Schema.org	LLM-LD Addition
`name`, `description`	`llmld:summary.one_liner`, `llmld:summary.key_facts`
`contactPoint`	`llmld:actions` with typed action endpoints
`hasOfferCatalog`	`llmld:products` with structured pricing
No equivalent	`llmld:context.ideal_customer`
No equivalent	`llmld:trust.reviews`

12. Security Considerations

Data Exposure

The llm-index.json file is publicly accessible. Do NOT include:

Private pricing (internal pricing, cost basis)
Competitive intelligence strategies
Personally identifiable information (PII) not already publicly available
Authentication credentials or API keys

Data Accuracy

Inaccurate data can mislead AI systems. Implement a process for keeping data current. Set llmld:meta.expires to indicate when data becomes stale.

Content Injection

AI systems consuming this data may pass it to language models. Avoid including content that could cause prompt injection or manipulation of downstream AI systems.

13. IANA Considerations

This specification requests registration of the llm-index link relation type.

Relation Name: llm-index

Description: Identifies a resource containing machine-readable AI-layer index data conforming to the LLM-LD specification.

Reference: This specification.

14. Complete Example

json

{
  "@context": ["https://schema.org", "https://llmld.org/v1"],
  "@type": "llmld:AIWebsite",
  "@id": "https://ai.acmedental.com/.well-known/llm-index.json",
  "llmld:meta": {
    "version": "1.0",
    "generated": "2026-02-01T12:00:00Z",
    "expires": "2026-03-01T12:00:00Z",
    "language": "en-US"
  },
  "llmld:site": {
    "name": "Acme Dental",
    "type": "Business",
    "industry": ["Healthcare", "Dentistry"],
    "description": "Family dental practice in Springfield, IL.",
    "founded": "1999",
    "domains": {
      "primary": "https://www.acmedental.com",
      "ai": "https://ai.acmedental.com"
    },
    "location": {
      "headquarters": {
        "address": "123 Main Street",
        "city": "Springfield",
        "state": "IL",
        "postal_code": "62701",
        "country": "US"
      },
      "service_area": "Springfield, IL and surrounding communities"
    }
  },
  "llmld:primaryEntity": {
    "@id": "#organization",
    "@type": "Dentist",
    "name": "Acme Dental",
    "url": "https://www.acmedental.com"
  },
  "llmld:summary": {
    "one_liner": "Family dental practice in downtown Springfield specializing in pediatric and cosmetic dentistry",
    "paragraph": "Acme Dental has been serving Springfield, IL since 1999. We offer a full range of dental services including pediatric dentistry, cosmetic procedures, and emergency care.",
    "key_facts": [
      "25+ years serving the Springfield community",
      "Certified pediatric dental specialists on staff",
      "Same-day emergency appointments available",
      "Accepts all major insurance plans"
    ],
    "target_customers": ["Families", "Children", "Adults seeking cosmetic dentistry"]
  },
  "llmld:actions": {
    "primary": [
      {
        "id": "book-appointment",
        "name": "Book an Appointment",
        "url": "https://www.acmedental.com/book",
        "type": "schedule",
        "priority": 1
      }
    ],
    "contact": [
      {
        "id": "call-us",
        "name": "Call Us",
        "type": "contact",
        "channels": [
          {"type": "phone", "value": "+1-217-555-1234", "label": "Main Line"}
        ]
      }
    ]
  },
  "llmld:contact": {
    "general": {
      "phone": "+1-217-555-1234",
      "email": "hello@acmedental.com",
      "hours": "Mon-Fri 8am-5pm, Sat 9am-2pm"
    },
    "address": {
      "street": "123 Main Street",
      "city": "Springfield",
      "state": "IL",
      "postal_code": "62701",
      "country": "US"
    }
  },
  "llmld:services": [
    {
      "@id": "#service-pediatric",
      "@type": "MedicalTherapy",
      "name": "Pediatric Dentistry",
      "description": "Complete dental care for children from infancy through adolescence.",
      "service_type": "recurring"
    }
  ],
  "llmld:verification": {
    "status": "verified",
    "verified_by": "https://llmdisco.com",
    "verified_at": "2026-02-01T00:00:00Z",
    "conformance": {"level": 3, "level_name": "Agent-Ready"}
  }
}

15. JSON Schema

A formal JSON Schema for validating llm-index.json files is maintained at:

https://llmld.org/schema/v1/llm-index.schema.json

16. Changelog

v1.0 (February 2026)

Initial draft specification
Three conformance levels defined
Core, entity, and agent intelligence property groups
Well-known URI placement
LLM Disco Directory integration

Guide Overview | Quick Start

LLM-LD 1.0 Specification Fresh ​

Abstract ​

Table of Contents ​

1. Introduction ​

1.1 Background ​

1.2 Goals ​

1.3 Audience ​

1.4 Document Conventions ​

2. Terminology ​

3. Conformance Levels ​

3.1 Level 1: Crawl-Ready ​

3.2 Level 2: Ingest-Ready ​

3.3 Level 3: Agent-Ready ​

3.4 Conformance Declaration ​

4. File Structure ​

4.1 File Format ​

4.2 Context Declaration ​

4.3 Property Namespacing ​

4.4 Size Considerations ​

5. Core Properties ​

5.1 llmld:meta ​

5.2 llmld:site ​

5.3 llmld:primaryEntity ​

5.4 llmld:summary ​

5.5 llmld:pages ​

5.6 llmld:actions ​

5.7 llmld:contact ​

6. Entity Properties ​

6.1 llmld:products ​

6.2 llmld:services ​

6.3 llmld:people ​

6.4 llmld:locations ​

6.5 llmld:content ​

7. Agent Intelligence Properties ​

7.1 llmld:context ​

7.2 llmld:faq ​

7.3 llmld:trust ​

8. Verification Properties ​

8.1 llmld:verification ​

9. Meta Properties ​

9.1 llmld:pointers ​

10. File Placement and Discovery ​

10.1 Well-Known URI (Recommended) ​

10.2 Root Path (Alternative) ​

10.3 Discovery via robots.txt ​

10.4 Discovery via HTML Link Element ​

10.5 HTTP Headers ​

10.6 robots.txt Declaration ​

11. Relationship to Schema.org ​

12. Security Considerations ​

Data Exposure ​

Data Accuracy ​

Content Injection ​

13. IANA Considerations ​

14. Complete Example ​

15. JSON Schema ​

16. Changelog ​

LLM-LD 1.0 Specification Fresh

Abstract

Table of Contents

1. Introduction

1.1 Background

1.2 Goals

1.3 Audience

1.4 Document Conventions

2. Terminology

3. Conformance Levels

3.1 Level 1: Crawl-Ready

3.2 Level 2: Ingest-Ready

3.3 Level 3: Agent-Ready

3.4 Conformance Declaration

4. File Structure

4.1 File Format

4.2 Context Declaration

4.3 Property Namespacing

4.4 Size Considerations

5. Core Properties

5.1 llmld:meta

5.2 llmld:site

5.3 llmld:primaryEntity

5.4 llmld:summary

5.5 llmld:pages

5.6 llmld:actions

5.7 llmld:contact

6. Entity Properties

6.1 llmld:products

6.2 llmld:services

6.3 llmld:people

6.4 llmld:locations

6.5 llmld:content

7. Agent Intelligence Properties

7.1 llmld:context

7.2 llmld:faq

7.3 llmld:trust

8. Verification Properties

8.1 llmld:verification

9. Meta Properties

9.1 llmld:pointers

10. File Placement and Discovery

10.1 Well-Known URI (Recommended)

10.2 Root Path (Alternative)

10.3 Discovery via robots.txt

10.4 Discovery via HTML Link Element

10.5 HTTP Headers

10.6 robots.txt Declaration

11. Relationship to Schema.org

12. Security Considerations

Data Exposure

Data Accuracy

Content Injection

13. IANA Considerations

14. Complete Example

15. JSON Schema

16. Changelog