AI Visibility Measurement Methodology
At RANKVERI, we believe no LLM visibility metric should ever be presented as an absolute truth or an opaque black box. Here is exactly how our algorithms calculate your scores.
1. Multi-Pass Sampling (N = 3 to 5)
Absorbing the natural non-determinism of AI models
Unlike traditional search engines (Google), generative AI models (ChatGPT, Claude, Gemini, Perplexity, Grok) can vary their responses slightly for the exact same query posed at different times.
To obtain a robust metric, RANKVERI queries each prompt 3 to 5 times per engine. A citation rate is not an arbitrary yes/no, but the exact proportion of responses recommending your brand.
2. 3-Pass Hybrid Detection Algorithm
Maximizing recall without false attribution
Strict matching of brand name, acronyms, and spelling variants.
Named entity recognition using natural language processing (spaCy).
Evaluation model arbitrating indirect and contextual brand mentions.
3. Immutable Proof via Raw Transcript
Access the full text of every captured AI response
Every measurement taken by RANKVERI is preserved in its original raw state (timestamp, exact engine text, detected position). You can click on any reading at any time to read the full robot response and verify the highlighted citation with your own eyes.