AI Protocols for Websites: Discovery, Meaning, Trust, and Handoff

1. Discovery tells AI where the useful public files and pages live.
2. Meaning uses Schema.org identities and Dataset definitions to explain what those resources contain.
3. Trust publishes health, scoring, relationship, and rendered-audit evidence that can be checked.
4. Handoff gives another crawler, assistant, developer, or coding agent enough context to continue without starting from a blank page.

A protocol becomes useful when each job has a clear owner. One file should not pretend to do all four.

Digital Karma Federation gives a website an ordered discovery system. The manifest identifies the site and points to the current files. The catalog names the real content families. The rendered receipt proves those Dataset identities appear across the canonical sitemap.

The health, score, and federation files then add current condition and relationship context. Root discovery files such as /llm.txt, /robots.txt, and /sitemap.xml help crawlers enter the system from addresses they already know to check.

/ai/catalog.json is the source of truth for the site's DataCatalog and Dataset identities. A site with articles, products, FAQs, and directory records publishes those as separate content families instead of hiding them under one generic library label.

Every canonical page reuses the same stable catalog IDs and connects its main subject to the matching Dataset. That gives an AI system the same answer whether it begins with the catalog or lands on an article first.

Digital Karma Federation is not a private messaging system, an authentication layer, a tool-calling protocol, or an application programming interface. It does not tell one AI assistant how to operate another service.

Its job is narrower and public: help a website identify itself, describe its reusable information, publish its relationships, and prove that the rendered pages match the catalog.

A crawler should not need one lucky visit to one forgotten file to understand the site. The same identity and Dataset map should remain available from the discovery layer and from the pages themselves.

That makes the setup easier to audit and easier to maintain. Start with the current standards when you need the exact rules, or use the live audit to see which parts a public site already exposes.