How AI Visibility works

AI Visibility tracks how often your brand, products, or pages get mentioned in answers from major LLMs when users ask buyer-relevant questions. It's the answer to "are we showing up in ChatGPT?" — at scale, daily, with history.

Which LLMs we query

• ChatGPT — both gpt-5 and gpt-4o; web-search-enabled and offline variants.
• Perplexity — sonar and sonar-large, with citation tracking.
• Google Gemini — gemini-2.5-pro and gemini-flash, search-grounded.
• Claude — sonnet-4.6 and opus-4.7, both with and without web search.

Each engine is queried independently — sharing infrastructure between them would bias the comparison. Results are de-duplicated per (engine, prompt, day) before being scored.

How often

Every prompt, every engine, every day. Higher-cadence is available on Enterprise (up to hourly for priority prompts). Results are stored permanently so you can chart your mention trajectory over months or years.

How the queries are made

Each query is a fresh, account-less, history-less request. No cached system prompts, no user context, no memory. The query is the prompt you set, verbatim. The response is whatever the LLM would say to a brand-new user asking that same question right now.

What gets stored

• Full response text from every query, every engine, every day.
• Extracted mentions (your brand, products, URLs, competitor mentions).
• Citation links (for engines that cite sources — Perplexity, Gemini with web).
• Sentiment classification on your mentions (positive / neutral / negative / cautionary).
• Day-over-day deltas surfaced in the change feed.

Share-of-voice math

For a given prompt, SoV = (responses where you appear in the recommended-options section) ÷ (total responses across all engines for that prompt). "Mentioned in passing" doesn't count — only structural recommendations do. This is conservative on purpose; it's the number that correlates with real downstream traffic.