AIVI supports decisions; it does not replace one.
The AI Visibility Index combines four bounded evidence groups. Every component is traceable to a file, response, policy rule, or documented probe. Missing evidence is marked unknown and does not receive an invented green pass.
| Component | Weight | What it measures | Reference |
|---|---|---|---|
| S_crawl | 40% | robots.txt allowance multiplied by observed live probe score for the core crawler subset. | Policy: Robots Layering → |
| S_infra | 20% | Presence and quality signals for llms.txt, llms-full.txt, sitemap, and AGENTS.md. | Policy: llms.txt Standard → |
| S_fingerprint | 15% | Detected bot-defense evidence, WAF challenge behavior, and Web Bot Auth verification. | Glossary: User-Agent & Headers → |
| S_geo | 25% | Markdown alternate, JSON-LD structured data, and content entity trust signals. | Glossary: Schema.org & JSON-LD → |
Formula
AIVI = 0.40·S_crawl + 0.20·S_infra + 0.15·S_fingerprint + 0.25·S_geo
Coverage and limitations
The registry contains 219+ catalogued bot definitions. A bounded core subset receives live User-Agent probes so the scanner remains respectful of subrequest and response budgets. A live status is an observation from the scan edge, not proof of vendor identity or a promise that a crawler will obey robots.txt. Sensitive content must still use authentication and authorization.
The score is most useful when compared with its own prior run. A low score identifies evidence gaps and conflicts to investigate; it does not rank the quality of your content or predict citations by itself.