The short answer
To track AI visibility across ChatGPT, Gemini, and Perplexity, run one shared set of category prompts against every engine on a schedule and read the results per engine rather than as a single blended number. Each engine retrieves and cites sources differently, so you can be strong in Perplexity and weak in Gemini at the same time, and only per-engine numbers reveal that. For each engine, record citation rate, share of model versus competitors, and sentiment, then track those over time.
Use one prompt set, read it per engine
The trick is to keep the input identical and the output separated. Define the questions a real buyer would ask in your category, then send that same set to every engine. Because the prompts are shared, the engines become comparable: any difference in your results is a difference in the engines, not the questions. You then read each engine on its own, since a crawler block, a missing citation, or weak authority can hurt you in one engine while leaving another untouched.
What to record for each engine
For every engine, capture the same three signals: citation rate (how often you appear), share of model (your slice versus competitors), and sentiment (how you are described). Running the set on a schedule turns each into a trend, so you can tell whether a content change moved ChatGPT, Gemini, and Perplexity together or just one of them. A single summary score on top is handy for reporting, but the per-engine detail is where you decide what to fix.
AEO Goal runs one prompt set across all these engines and reports them side by side. See AI citation tracking for the per-engine view, or competitor AI visibility to measure share of model against rivals. To start with a baseline, run the free AI visibility scan.