Track your brand, engine by engine

Each engine grounds its answers through a different API mechanism and cites a different mix of sources, so the same brand can appear in three answers out of five. These guides cover what that means on each engine and how to measure it rather than spot-check it. The shared concepts live in Learn.

How does each engine ground its answers?

All five run live web searches when asked to, but each exposes it differently and returns citations in its own shape. Reading citations from that structured metadata — rather than parsing URLs out of the answer text — is what separates a real measurement from a plausible one.

Grounding mechanism per engine, as used by AI Visibility Tracker
EngineHow it is groundedGuide
ChatGPTOpenAI Responses API with the web_search tool; url_citation annotationsTrack your brand in ChatGPT
ClaudeAnthropic web_search server tool; structured result blocksTrack your brand in Claude
Geminigoogle_search tool; groundingMetadata.groundingChunksTrack your brand in Gemini
Perplexitysonar models — search is always on, no ungrounded modeTrack your brand in Perplexity
GrokRouted via OpenRouter with the :online suffixTrack your brand in Grok

Which engines should you actually track?

ChatGPT always, because it has the largest reach of any assistant. Gemini if organic search matters to you, since it is the closest available proxy for Google’s AI surfaces. Perplexity because it is the fastest feedback loop and the densest source of competitor data. Claude if your buyers are technical. Grok only if your audience is genuinely on X.

Adding an engine multiplies your daily result count and your API bill by the same factor, so a fifth engine you do not need is not free. The reasoning behind each of those calls is in the individual guides, and the metric they all feed is Share of AI Answer.

Frequently asked questions

Can I track all five AI engines with one API key?

Yes. A single OpenRouter key can carry every engine, with web grounding applied automatically. You can also use each vendor’s own key directly — OpenAI, Anthropic, Google or Perplexity — and AI Visibility Tracker will route each engine through whichever key you have configured for it.

How many prompts should I track per engine?

The same prompt set across every engine — that is what makes the per-engine numbers comparable. Ten to fifteen core prompts run daily is enough to start; across five engines that is 75 results per run and roughly 500 a week, which is enough to distinguish a real shift from ordinary variance.