Agntbase
Playbook

AI readiness playbook: website to canonical profile

Teams usually start with a website that is human-readable but inconsistent for AI systems. The fastest path is not adding random files first. The right order is: parse facts, approve draft, publish canonical profile in hub, then place minimal package on client site.

Execution order that works

  1. Run readiness check and capture hard blockers.
  2. Build approval draft with confirmed facts and numbered gaps.
  3. Publish full canonical profile in hub (`/p/{slug}/`).
  4. Generate minimal website package (`company-profile.json`, `agenthub.json`, `llms.txt`).
  5. Re-check and measure score growth after baseline and after expert pass.

This keeps one source of truth in hub and avoids overloading client sites with full governance payloads.

Research references

  1. Agentic Web Interface (arXiv:2506.10953)
  2. Permissions manifest layer (arXiv:2601.02371)
  3. Structured linked data memory layer (arXiv:2603.10700)
  4. Google Search Central: succeeding in AI Search
  5. McKinsey: New front door to the internet (AI search)
Run free check now Open registry Start setup flow