The Signed Web: Why AI Needs Verified Knowledge, and How We Built It

AI Is Guessing About Your Business

Ask any AI assistant about your company. Go ahead — try it right now. Ask ChatGPT, Claude, or Gemini: "What does [your company] do?"

What you get back is a guess. The AI scraped your HTML, parsed your marketing copy, maybe found a Wikipedia entry or a LinkedIn page, and assembled its best interpretation. Sometimes it's close. Sometimes it confidently tells people you do things you stopped doing three years ago. Sometimes it invents products you never made.

This is the state of the art in 2026. Billions of dollars in AI infrastructure, and the fundamental question — "What does this organization actually do?" — gets answered by web scraping.

We built something better.

What We Built

The AI Discovery Standard is a simple proposal: put a structured JSON document at /.well-known/ai on your website that tells AI agents exactly who you are, what you do, what you allow, and how to verify all of it.

Think of it as the difference between someone reading your business card versus rummaging through your office and guessing.

But we didn't just write a specification and hope someone would implement it. We built the first implementation ourselves — a free WordPress plugin that any site owner can install in minutes.

The First Implementation: A WordPress Plugin

The Rootz AI Discovery Plugin for WordPress does three things:

1. Creates your AI identity — generates a structured ai.json manifest with your organization's name, mission, products, policies, and contact information
2. Hashes your content — computes SHA-256 hashes for every published page so AI agents can verify content hasn't been tampered with
3. Signs everything — generates a secp256k1 cryptographic wallet for your site and signs the entire manifest, proving who published the data

The plugin serves the manifest at /.well-known/ai — the same convention used by SSL certificates (/.well-known/acme-challenge) and security policies (/.well-known/security.txt). It's a standard location where machines know to look for machine-readable instructions.

Why Signing Matters: Lessons from Trusted Computing

For those of us who spent decades in trusted computing, the signing part is where it gets interesting.

Every other AI discovery approach serves unsigned data. An attacker who compromises a web server can serve false information to AI agents — and the AI has no way to know. It's the same problem that motivated the Trusted Platform Module twenty years ago: without a measurement and attestation mechanism, you're trusting the platform without verifying it.

The AI Discovery Standard applies the same principle to web content:

An AI agent reading a signed manifest doesn't just get your data — it gets a cryptographic assertion that the data hasn't been tampered with since the site operator signed it. It can verify WHO published the data, not just WHERE it was served from.

We call this principle Origin, not Trust. The same principle that made measured boot work for platform integrity can make the web work for AI integrity.

What the Signature Looks Like

Every manifest includes a signature block:

"_signature": {
  "digitalName": "0xD36AAf65a91bB7dc69942cF6B6d1dBa4Ef171664",
  "network": "polygon",
  "contentHash": "sha256:a1b2c3d4...",
  "signedAt": "2026-02-27T00:00:00Z",
  "method": "epistery-domain-v1"
}

And every page gets an individual content hash:

"pages": [
  {
    "path": "/technology",
    "contentHash": "sha256:7f3a9b2e...",
    "purpose": "technical documentation"
  }
]

An AI agent can fetch any page, hash it, and compare. If the hash matches, the content is exactly what the site operator signed. If it doesn't match, something changed. This is verification any machine can perform, instantly, without trusting anyone.

The Demo That Proves It: verifyPageHash

The plugin exposes a tool called verifyPageHash that demonstrates the entire concept in a single API call. An AI agent sends a URL; the plugin fetches the page, hashes it, and compares the hash against the signed manifest. The response tells the AI whether the content is verified or has been modified.

This is the simplest possible proof of content integrity. No blockchain lookup required, no complex verification chain — just cryptographic math that any machine can execute. And it works right now, on any WordPress site running the plugin.

Testing Against the Real World

We tested the standard against real websites to see how AI agents fare today — with and without structured data.

The gap is striking. With structured data, an AI agent knows exactly what the organization does, what it allows, and how to cite it — with cryptographic proof. Without it, the AI is guessing, and two-thirds of real websites can't even tell AI whether quoting is permitted.

Three Tiers of Knowledge

The standard uses a three-tier architecture that scales from a simple business card to a complete organizational encyclopedia:

Early Adopters

Within days of publishing the standard, inblock.io (the Aqua Protocol team) implemented the full three-tier architecture and scored Grade A (95/100) on our AI readability scanner. They described the standard as providing "verifiable, structured data rather than blind faith in platforms" — language that resonated because it captures exactly what's missing from the current web.

Unveiled Muse, a wedding inspiration platform built by Lexi, has also implemented the standard — proving it works beyond the blockchain and security communities. A consumer site helping couples plan weddings now speaks AI natively, with structured knowledge about budget tiers, categories, and photographer credits that no scraper would ever extract correctly.

Made in USA One (OTC: USDW) and its certification arm Verity One Ltd are adopting the AI Discovery Standard for enterprise supply chain verification. With 22 years of Made in USA certification experience, they're bringing the standard to a sector where origin and authenticity are the entire business — using /.well-known/ai to let AI agents verify product provenance and certification status directly from the source.

The pattern is clear: from blockchain infrastructure to wedding planning to supply chain certification, the standard works anywhere an organization wants AI to get its facts right. Six sites now publish at /.well-known/ai. The standard is twelve days old.

What This Means for Your Business

Every day you don't have /.well-known/ai, AI agents are making up answers about your organization. They're guessing at your products, inventing your policies, and presenting their interpretations as fact to the people asking about you.

Adding the standard changes the power dynamic:

• Accuracy: AI responses about your organization come from you, not from scraped interpretations
• Currency: Update your manifest and AI knows immediately — no waiting for a crawler to re-index
• Control: You define your quoting policy, your attribution requirements, your contact channels
• Verification: Anyone can check the signature and confirm the data came from you
• Visibility: Brands cited by AI see 35% more organic clicks and 91% more paid clicks (Seer Interactive, 2024)

This isn't theoretical. It's working right now. Point any AI agent at rootz.global/.well-known/ai and ask it about us. Then try the same with any company that doesn't have the standard.

Get Started

You have three paths:

An Open Standard

The AI Discovery Standard is published under CC-BY-4.0. Anyone can implement it, on any platform, for free. The specification is at /ai/standard-v1.2.md. We're working toward RFC 8615 registration of the /.well-known/ai URI.

The WordPress plugin is GPL v2. The scanner is free. The generator is free. The standard itself is Creative Commons. We want every website on earth to have a /.well-known/ai file — that's the goal, not vendor lock-in.

Because the web doesn't need another walled garden. It needs infrastructure. Infrastructure that makes AI accurate, verifiable, and accountable to the organizations it represents.

The Bigger Picture

The AI Discovery Standard is the first visible application of what we've been building at Rootz: infrastructure where data carries proof of its own origin and ownership. We call this Data with Origin and Ownership — the idea that every piece of digital information should be able to prove where it came from, who created it, and whether it's been modified.

For websites, that means a signed manifest at /.well-known/ai. For supply chains, it means Data Wallets with molecular-level physical fingerprints. For AI conversations, it means archived thinking with cryptographic proof of origin. The underlying principle is the same: don't trust — verify. And give the tools to verify to everyone, not just the gatekeepers.

The web was built for humans. We're adding a layer for machines — not to replace the human web, but to let AI understand it accurately, directly, and verifiably.

Stop letting scrapers decide what your company means. Tell AI yourself.