Why AI Discovery
The Problem
When an AI agent visits your website today, it faces a series of unanswerable questions:
- What does this organization actually do?
- Am I allowed to quote this content?
- Can I use this for training?
- What policies apply to automated access?
- Who signed this content and can I verify it?
- What pages exist beyond what I can crawl?
Without structured data, AI agents are forced to scrape HTML, parse legal documents written for humans, and guess at permissions. The results are unreliable, expensive, and often wrong.
We Tested It
In February 2026, we ran a controlled experiment. We asked AI agents to answer 10 standard questions about four websites — one with /.well-known/ai and three without. The results were stark:
| Site | Has AI Discovery | HIGH Confidence | LOW Confidence |
|---|---|---|---|
| discover.rootz.global | Yes | 9 / 10 | 0 |
| Intel | No | 7 / 10 | 0 |
| Naoris Protocol | No | 6 / 10 | 2 |
| AdventHealth | No | 4 / 10 | 4 |
The site with AI Discovery scored 9 out of 10 with HIGH confidence. AdventHealth — a major healthcare system with 92,000 employees — scored just 4, with policy pages that returned 404 errors or redirected to the wrong subdomain.
The Permission Crisis
The most critical finding was about content licensing. Only one site could give AI a machine-readable answer to “Can I quote this content?”
- With AI Discovery:
permissions: ["quote", "summarize", "cache_24h"]— instant, unambiguous - Intel: Prohibition buried in Terms of Use legal text
- AdventHealth: Unknown — policy pages broken
- Naoris Protocol: Unknown — terms not retrievable
Two out of three real-world sites cannot even tell an AI agent whether it is allowed to read them.
What AI Discovery Provides
The AI Discovery Standard creates a structured interface between your website and AI agents. A single endpoint at /.well-known/ai provides:
Organization Identity
Name, domain, mission, sector, legal name, headquarters — structured JSON that any AI can parse without scraping your About page.
Content Policies
Machine-readable permissions and restrictions. Not legal text to interpret — explicit arrays: ["quote", "summarize"], ["no_training"]. AI agents know immediately what they can and cannot do.
Cryptographic Signing
Every response is signed with a secp256k1 key. AI agents can verify the content hash, check the signer address, and confirm the data has not been tampered with. This is content attestation — proof of origin that works without trusting the transport layer.
Knowledge Base
Structured information about your organization — about text, products, glossary terms — available in one API call instead of crawling dozens of pages.
AI-Optimized Feed
Recent content formatted for AI consumption. Titles, summaries, categories, and publication dates without the overhead of parsing HTML blog layouts.
Contact Information
Operator name, general email, AI-specific support email, privacy contact — so AI agent developers know exactly who to reach.
The Numbers
Structured path: 4 API calls, ~3,200 tokens of input data, zero failed requests.
Scraping path: 4-8 page fetches, ~5,000+ tokens per site, frequent failures and redirects.
AI Discovery is faster, cheaper, more reliable, and produces better results.
Get Started
The Rootz AI Discovery WordPress Plugin adds /.well-known/ai to any WordPress site in minutes. The specification is open (CC-BY-4.0) and works on any platform.