The AI Visibility Index · Methodology
How we measure visibility for AI SEO decisions.
AI visibility is the share of repeated category buyer-question checks where an assistant names a company. Answers vary, so we publish a 95% confidence range and the observed mention count, not a single exact verdict. It is measured, not modeled. It can guide what a B2B team investigates next, but it does not measure revenue or prove that an AI SEO action caused a commercial result.
The definition
observed rate = mentions ÷ completed checks
A check is one completed answer to one buyer question from one assistant. We repeat each question three times. A mention is a check where the answer names the company. The visible range is a 95% Wilson interval around that observed rate, which stays appropriately wide for small samples.
How we measure it
- Real buyer questions. For each category we ask questions a real buyer would use when shortlisting, phrased naturally and never naming a measured company.
- Shared, repeated answers. Each question is sampled three times per enabled assistant. Every company is checked against the exact same completed answers, so one company cannot get an easier set of responses than another.
- Real provider responses with web search. Questions go to the configured OpenAI, Perplexity, Anthropic, and Google models with their web-search tools. Only completed responses enter the denominator.
- Deterministic mention detection. A company counts as named only on an exact match of its name or domain, never a loose substring, so a short name cannot match by accident and a miss can never be faked into a hit.
- We count, we do not judge. Being named does not necessarily mean endorsement. We also record which competitors appear in the same real answers.
What it is, and what it is not
- It is a sampled snapshot of public answers on the measurement date. The range describes sampling uncertainty, not a promise about every future answer.
- It measures being named by AI, not endorsement, traffic, revenue, or product quality.
- It is measured on public data about public companies. No private or customer data is used.
Reproduce it
Open ChatGPT or Perplexity, ask the same category questions, and see who gets named. That is the whole method. Want an AI SEO baseline measured on your buyer questions? Run the free AI SEO scan.
Method version v2 · shared answers · three samples · Wilson 95% range · The AI Visibility Index by Eli