01
A useful definition
AI Visibility is the observed presence and treatment of a brand in answers produced by defined AI systems for a defined set of scenarios. It is not a permanent property of a company and it is not one universal score.
A defensible result names the system or surface, exact model where applicable, language, region, prompt family, repeat count, date and whether web search was available. Without that configuration, a percentage is difficult to interpret or reproduce.
02
Readiness and visibility answer different questions
Public Readiness asks whether machines can reach, parse and reuse information on a website. It can inspect HTTP access, robots rules, canonical URLs, structured data, entity clarity, content structure, contact consistency and block-level citability heuristics.
Real AI Visibility asks what selected AI systems actually answered. That requires a separate measurement cycle and an Evidence Ledger. Improving readiness may make a site clearer, but it does not prove that ChatGPT, Gemini or Perplexity mentioned or recommended the brand.
- Readiness evidence comes from the website and deterministic rules.
- Visibility evidence comes from dated AI answers collected under a Configuration Lock.
- The two can be compared side by side, but should never be merged into an opaque composite score.
03
Visitor View and API View are not interchangeable
A consumer-facing answer surface with search available is a different channel from a direct API response produced by a fixed model with web search off. Selena reports Visitor View and API View separately, then calculates divergence only between results that are genuinely comparable.
This prevents a common mistake: presenting one API model response as everything a product or company 'knows'. It is only one response under one configuration.
04
What a sound measurement should preserve
The evidence needs both the answer and its context: prompt, system, model or surface, search status, language, region, repeat index, timestamps, citations and validation state. Cardinality must be planned before a run so missing or extra answers are visible rather than silently averaged away.
The result is decision support. It can show where a brand appears, which competitors appear instead, which sources are cited and what should be investigated next. It cannot guarantee future rankings or recommendations.