How the Agent Readiness Score works.
Three pillars. Twenty-plus deterministic checks. A site-type-aware weighting model so a SaaS dashboard isn’t judged against a content publisher.
Can AI agents reach and parse your site? Covers robots.txt + directives, sitemap.xml, AI-bot allow/deny rules (GPTBot, ClaudeBot, PerplexityBot …), Cloudflare content signals, /llms.txt, /.well-known/mcp.json (Anthropic MCP), /.well-known/agent-skills, /.well-known/http-message-signatures-directory (web-bot auth), commerce protocols (ACP/UCP/MPP/x402), Markdown content negotiation, and WAF agent-blocking.
Can an agent complete the canonical jobs for your site type? Probes find-info, signup/login, contact/demo, cart/checkout paths. Grades CTA clarity and homepage value-prop legibility.
Will agents cite and recommend you? Covers Schema.org coverage and entity consistency, semantic heading hierarchy, alt-text density, freshness signals, trust links (Privacy/Terms/About/Contact), and homepage content density.
The pillar weights are tuned to the business archetype. A docs/developer site is judged 45% Technical; a content publisher is 50% Recommendation. The classifier looks at homepage copy, link text, Schema.org types, and canonical task paths to assign a type with a confidence score.
| Archetype | Technical | Task | Recommendation |
|---|---|---|---|
| Content publisher | 30% | 20% | 50% |
| Commerce | 30% | 35% | 35% |
| Lead generation | 25% | 40% | 35% |
| SaaS / product | 30% | 35% | 35% |
| Docs / developer | 45% | 30% | 25% |
| Local business | 20% | 40% | 40% |
| Campaign microsite | 25% | 35% | 40% |
| Hybrid | 30% | 35% | 35% |