Specification v1.2 - AI Discovery Standard

Abstract

This specification defines a standard mechanism for websites to present structured, machine-readable organizational identity to AI agents via the well-known URI /.well-known/ai. It enables AI agents to discover, understand, and verify organizational information directly from the authoritative source — without scraping, without intermediaries, and with cryptographic proof of origin.

Note: The key words MUST, SHOULD, MAY in this document are to be interpreted as described in RFC 2119.

1. Problem Statement

In 1995, publishers scanned magazines to put them on the web. The real revolution came when content was structured for the medium. Today, companies are making the same mistake with AI: they build websites for human browsers, then let scrapers interpret their HTML.

The result:

AI agents hallucinate company details because they guess from HTML
No mechanism to verify if AI-presented information is accurate or current
Organizations have zero control over how AI represents them
Scrapers aggregate and repackage without attribution or verification

Existing standards address parts of this but not the core problem:

Standard	Purpose	Gap
`robots.txt`	What can I crawl?	Says nothing about what to understand
`agents.json`	What APIs can agents call?	Agent capabilities, not company knowledge
`a2a/agent.json`	How do agents communicate?	Agent-to-agent protocol, not knowledge
`ai-plugin.json`	ChatGPT plugin manifest	Plugin functionality, not organizational identity
Schema.org JSON-LD	Page-level structured data	Per-page, no single discovery endpoint

None answer the fundamental question: "What does this organization do, and how can I verify it?"

2. The Solution: `/.well-known/ai`

A single JSON document at a well-known URI that tells any AI agent:

Who the organization is (name, people, Digital Name)
What it does (products, applications, core concepts)
Where to find deep knowledge (knowledge endpoint, feed endpoint)
How to verify the information is authentic (cryptographic signature)

2.1 Three-Tier Architecture

/.well-known/ai              Discovery (who, what, where to learn more)
    |
    +-- /ai/knowledge.json    Knowledge (deep organizational encyclopedia)
    |
    +-- /ai/feed.json         Feed (news, updates, changes)
    |
    +-- /ai/content           Content (full page/post access)
    |
    +-- /ai/policies          Policies (licensing, AI training, usage rights)

Tier 1: Discovery (/.well-known/ai) — Compact organizational profile with links to deeper endpoints. Includes identity, pages, core concepts, and cryptographic signature.

Tier 2: Knowledge — Full organizational encyclopedia with detailed glossary, products, team, market analysis, and technology stack.

Tier 3: Feed — AI-optimized chronological updates with structured items, key facts, tags, and related concepts.

3. File Format Specification

3.1 Discovery File (`/.well-known/ai`)

Content-Type: application/json
CORS: Access-Control-Allow-Origin: *
Cache: Recommended 24 hours

Required Fields

Field	Type	Description
`$schema`	string	Schema URL for validation
`specVersion`	string	Specification version (semver). Current: `1.2.0`
`standard`	string	MUST be `"rootz-ai-discovery"`
`generated`	string	ISO 8601 generation timestamp
`organization`	object	Organization identity (see 3.1.1)
`coreConcepts`	array	Glossary of key terms

3.1.1 Organization Object

Field	Type	Required	Description
`name`	string	Yes	Organization display name
`domain`	string	Yes	Primary domain
`mission`	string	Yes	Mission statement
`sector`	array	Yes	Industry sectors
`legalName`	string	No	Legal entity name
`digitalName`	string	No	Blockchain address (Digital Name)
`founded`	string	No	Year founded
`headquarters`	string	No	Location
`tagline`	string	No	Short tagline
`blockchain`	string	No	Blockchain network name

Optional Top-Level Fields

Field	Type	Description
`people`	array	Key team members
`knowledge`	object	Link to knowledge endpoint
`feed`	object	Link to feed endpoint
`content`	object	Link to content endpoint
`policies`	object	Link to policies endpoint
`pages`	array	Site map with semantic purpose and content hashes
`applications`	array	Products and services
`partners`	array	Strategic partnerships
`contact`	object	Contact information (operator, AI support, privacy)
`generator`	object	Tool that generated the manifest
`_signature`	object	Cryptographic signature block

3.2 Knowledge File

Content-Type: application/json · CORS: * · Cache: 1 hour recommended

Structure is flexible but SHOULD include:

summary — oneLiner, elevator, forAI paragraphs
glossary — term, definition, synonyms, category
products — name, description, status, features
team — name, role, bio, expertise
market — size, thesis, sectors
technology — blockchain, security, architecture

3.3 Feed File

Content-Type: application/json · CORS: * · Cache: 30 minutes recommended

Each item MUST include: id, title, published, summary. SHOULD include: url, category, tags, keyFacts, relatedConcepts.

3.4 Content Endpoint

Path: /.well-known/ai/content
Serves pages, posts, media, and custom content types as structured JSON. Each item includes an assertionType field: factual, editorial, or creative-work.

Segmented access: /content/pages, /content/posts, /content/media. Optional ?full_text=1 parameter for complete content.

3.5 Policies Endpoint

Path: /.well-known/ai/policies
Machine-readable content licensing, AI training permissions, quoting rights, and usage policies. Includes 100-word policy summaries optimized for AI context windows.

4. Cryptographic Verification

4.1 The Signature Block

Each JSON file MAY include a _signature block:

{
  "_signature": {
    "digitalName": "0xD36AAf65a91bB7dc69942cF6B6d1dBa4Ef171664",
    "network": "polygon",
    "contentHash": "sha256:a1b2c3d4e5f6...",
    "signedAt": "2026-02-15T00:00:00Z",
    "method": "content-hash"
  }
}

Field	Description
`digitalName`	Blockchain address of the signing authority
`network`	Blockchain network (e.g., `polygon`)
`contentHash`	SHA-256 hash of all fields excluding `_signature`
`signedAt`	ISO 8601 timestamp of signing
`method`	Signing method identifier

4.2 Verification Process

Extract the _signature block from the JSON
Compute SHA-256 hash of the remaining JSON (canonicalized, sorted keys)
Compare with contentHash — this verifies integrity
Optionally verify the digitalName on-chain to confirm organizational authority

4.3 Per-Page Content Hashes

The pages array MAY include a contentHash for each page entry. This is the SHA-256 hash of the page's text content, enabling AI agents to verify that the page content hasn't changed since the manifest was signed.

{
  "pages": [
    {
      "url": "/about",
      "title": "About Us",
      "purpose": "Company history and mission",
      "contentHash": "sha256:7f83b1657ff1fc53b92dc18148a1d65dfc2d4b1fa3d677284addd200126d9069"
    }
  ]
}

5. WebMCP Tools

The standard defines browser-based tools that AI assistants can invoke via the WebMCP protocol:

Tool	Category	Description
`getOrganizationInfo`	Discovery	Returns organization identity and overview
`getPolicies`	Discovery	Returns content licensing and AI policies
`getKnowledge`	Discovery	Returns deep organizational knowledge
`getFeed`	Discovery	Returns recent news and updates
`searchContent`	Actions	Search site content by keyword
`verifyPageHash`	Actions	Verify a page's content against its signed hash
`getStatus`	Meta	Returns implementation completeness score

6. HTML Integration

Implementations SHOULD include a link tag in the HTML <head>:

<link rel="ai-discovery" type="application/json" href="/.well-known/ai"
      title="AI Discovery Standard">

And an HTTP Link header:

Link: </.well-known/ai>; rel="ai-discovery"; type="application/json"

7. Relationship to Existing Standards

/.well-known/ai is complementary, not competing:

robots.txt tells crawlers what to index → ai.json tells AI agents what to understand
Schema.org JSON-LD provides per-page structured data → ai.json provides organization-wide discovery
agents.json / A2A defines agent capabilities → ai.json defines organizational knowledge
RSS/Atom provides chronological feeds → feed.json provides AI-optimized structured feeds

8. IANA Registration

This specification proposes registration of the well-known URI ai per RFC 8615:

Field	Value
URI suffix	`ai`
Change controller	AI Discovery Standard Steering Committee
Specification document	This document
Related information	Reference implementation at rootz.global

9. Implementation Guide

Minimum Viable Implementation

{
  "$schema": "https://rootz.global/ai/schemas/ai-discovery-v1.json",
  "specVersion": "1.2.0",
  "standard": "rootz-ai-discovery",
  "generated": "2026-02-27T00:00:00Z",
  "organization": {
    "name": "Example Corp",
    "domain": "example.com",
    "mission": "We build widgets that make the world better.",
    "sector": ["manufacturing", "technology"]
  },
  "coreConcepts": [
    {
      "term": "Widget",
      "definition": "Our core product: a self-assembling micro-component."
    }
  ]
}

Getting Started: A minimum implementation is 10 lines of JSON served at /.well-known/ai. For WordPress sites, the Rootz AI Discovery plugin generates the complete three-tier implementation automatically.

Reference Implementations

rootz.global — Reference implementation (static JSON + Express.js)
Rootz AI Discovery Plugin — WordPress plugin generating all tiers automatically
Veritize™ by Made in USA Inc. — Enterprise managed implementation

See Implementations for the full list and getting started guides.

References

RFC 8615 — Well-Known Uniform Resource Identifiers
RFC 2119 — Key words for use in RFCs
Schema.org — Structured data vocabulary
Google A2A — Agent-to-Agent protocol
CC-BY-4.0 — Creative Commons Attribution 4.0

License: This specification is published under the Creative Commons Attribution 4.0 International License (CC-BY-4.0).
First implementation: rootz.global
Published by: Rootz Corp — Digital Name 0xD36AAf65a91bB7dc69942cF6B6d1dBa4Ef171664

AI Discovery Standard

Abstract

1. Problem Statement

2. The Solution: /.well-known/ai

2.1 Three-Tier Architecture

3. File Format Specification

3.1 Discovery File (/.well-known/ai)

Required Fields

3.1.1 Organization Object

Optional Top-Level Fields

3.2 Knowledge File

3.3 Feed File

3.4 Content Endpoint

3.5 Policies Endpoint

4. Cryptographic Verification

4.1 The Signature Block

4.2 Verification Process

4.3 Per-Page Content Hashes

5. WebMCP Tools

6. HTML Integration

7. Relationship to Existing Standards

8. IANA Registration

9. Implementation Guide

Minimum Viable Implementation

Reference Implementations

References

2. The Solution: `/.well-known/ai`

3.1 Discovery File (`/.well-known/ai`)