How Get AI Vizible measures AI visibility (methodology)
By Get AI Vizible · Updated
Short answer
Get AI Vizible sends each brand's buyer questions to six AI engines (ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek) through their official APIs on a schedule, with the same neutral instruction for every engine. It then detects in code whether the brand and each tracked competitor is named, in what order, and whether the brand's website is cited, and uses a separate AI model to rate sentiment and check claims against facts the brand provides. Metrics are percentages over all answers in a scan.
1. Questions
Each brand has a fixed list of buyer questions (10 to 75 depending on plan), written in the words buyers use. Keeping the list stable makes scans comparable over time.
2. Engines
| Engine | How we query it | Web search |
|---|---|---|
| ChatGPT | OpenAI API, the model currently serving ChatGPT | No |
| Claude | Anthropic API | No |
| Gemini | Google Gemini API | No |
| Perplexity | Perplexity API (Sonar) | Yes, with citations |
| Grok | xAI model via Perplexity's Agent API | No |
| DeepSeek | DeepSeek model via Perplexity's Agent API | No |
Every engine receives the same instruction: answer the user's question, recommend specific brands by name where relevant, and cite source URLs when known. Most engines therefore answer from model knowledge, and Perplexity shows what live web search returns. Consumer apps can differ from API answers because of personalisation and search settings; API answers give a consistent baseline.
3. Analysis
- Mention: the brand or competitor name appears in the answer.
- Position: the order of first mention among the tracked brands (1 = named first).
- Citation: the brand's domain appears in the answer's cited URLs or text.
- Sources: every cited domain is counted and labelled as yours, a competitor's or third-party.
- Sentiment and FactCheck: for answers that mention the brand, a separate model rates the tone and checks each factual claim against the brand facts the customer provides, marking it accurate, inaccurate or unverifiable.
4. Metrics
| Metric | Definition |
|---|---|
| Visibility | % of answers that mention the brand |
| Average position | Mean position when mentioned |
| Share of voice | Brand mentions ÷ mentions of brand and all tracked competitors |
| Citation rate | % of answers citing the brand's domain |
| Sentiment score | Positive minus negative share of rated mentions |
All are calculated per scan, per engine and per question, and trended over time. See AI visibility metrics explained.
5. Schedule
Automatic scans run weekly on Starter, every 4 days on Growth and on a per-brand schedule on Agency. Customers can also run manual scans within their plan's daily limit.
6. Limits
- AI answers vary between runs; we report trends across many answers, not single answers.
- Mention detection matches names; brands with very generic names can produce false matches, so tracked names should be distinctive.
- Engines and models change; we record the model used for every answer.
Related
- AI visibility metrics explained: visibility, position, share of voice, citations and sentimentDefinitions and formulas for the five metrics that measure how AI assistants see your brand, with examples and how to improve each one.
- How to track brand mentions in ChatGPT, Perplexity, Gemini, Claude, Grok and DeepSeekManual and automated ways to track how AI assistants mention your brand and competitors, what to record, how often, and how to read the results.
- AI search and AEO glossary: 25 terms explainedPlain definitions of AEO, GEO, LLMO, AI visibility, share of voice, citation rate, AI Overviews, llms.txt, AI crawlers and more.