Specification v1.2 · CC-BY-4.0

The AI Discovery Standard

An open specification for how the web communicates with AI

Every website needs a way to tell AI agents who it is, what it does, and how to verify the answer. The AI Discovery Standard puts structured, signed organizational identity at /.well-known/ai — one URL, one format, cryptographically verifiable.

AI Can't Read Your Website

50 billion AI crawler requests hit the web every day. When they arrive at your domain, they scrape HTML designed for human eyes and guess what your organization does. The result: hallucinated facts, misattributed products, and zero accountability.

$67.4B
Lost to AI hallucinations in 2024
AllAboutAI 2024
33-51%
Hallucination rate for company & person queries
OpenAI o3 benchmarks
50B
AI crawler requests per day
Cloudflare 2025
200M+
Active websites with no AI interface
Netcraft

The web has a standard for search engines (robots.txt), for security (security.txt), for identity (openid-configuration). It has no standard for AI. Until now.

Three Tiers. One Standard.

Tier 1 — Discovery

Who You Are

/.well-known/ai

Organizational identity, mission, team, contact, products, and a glossary of core concepts. This is the business card AI reads first. Cryptographically signed with a content hash and Digital Name.

Tier 2 — Knowledge

What You Know

/.well-known/ai/knowledge

Deep organizational encyclopedia: detailed products, technology stack, market analysis, team expertise. Structured for AI agents that need comprehensive answers, not summaries.

Tier 3 — Feed

What's New

/.well-known/ai/feed

AI-optimized news feed with structured items, key facts, tags, and related concepts. Not RSS — purpose-built for AI agents that need to stay current on your organization.

Plus optional endpoints: /.well-known/ai/content for full content access, /.well-known/ai/policies for machine-readable licensing, and WebMCP tools for browser-based AI assistants.

Founding Members

The organizations shaping the standard. Founding member enrollment is open for a limited time.

Rootz
Standard Author & Reference Implementation
Made in USA
Enterprise & Government Certification
Your Logo Here
Founding Member Slot
Your Logo Here
Founding Member Slot

Founding members help shape the specification, participate in governance, and get early implementation support. Learn about membership →

Design Principles

Origin, Not Trust

Like a Trusted Platform Module measures what code is running rather than deciding if it's trusted, the AI Discovery Standard measures what a domain declares. AI agents verify origin; users decide trust.

Open by Default

The specification is CC-BY-4.0. Anyone can implement it. No licensing fees, no API keys, no vendor lock-in. The standard succeeds when every website has one, not when one company controls it.

Cryptographic Integrity

Every manifest includes a SHA-256 content hash and optional Digital Name signature. AI agents can verify that content hasn't been tampered with and trace it to an organizational identity on-chain.

Progressive Complexity

A minimum implementation is 10 lines of JSON. Full implementation adds knowledge, feed, content, policies, signing, and WebMCP tools. Start simple, grow as needed.

Domain Authority

The manifest lives on your domain at a well-known URI (RFC 8615). No intermediary hosts your identity data. Your domain, your declaration, your authority.

Multi-Vendor Governance

No single company controls the standard. Governance is designed for steering committee oversight, open working groups, and community participation. See the governance charter.

The Web Needs a Standard for AI

200 million websites. 50 billion daily AI requests. Zero standard protocol. The AI Discovery Standard changes that. Join us.