The short answer
You measure AI search visibility by running the real questions your buyers ask through the AI answer engines (ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews) and recording how often each engine mentions or cites your brand. The core numbers are citation rate (how often you appear), share of model (your slice of the answers versus competitors), and sentiment (how you are described). You run the same prompt set on a schedule so the numbers are comparable over time, then look at which engines cite you, which pages they pull from, and which competitors show up when you do not.
Start with the right prompts
Visibility is only meaningful against the questions your buyers actually ask. Build a prompt set from your category: the problems people describe, the comparisons they run, and the direct product questions. Keep it stable so you can compare week to week. A drifting prompt set makes the numbers look like they moved when really the questions changed.
Record three numbers per engine
For each prompt and each engine, capture whether you were mentioned or cited (citation rate), your presence relative to competitors in the same answer (share of model), and whether the mention was positive, neutral, or negative (sentiment). Engines differ, so keep the numbers per engine rather than averaging them into one figure that hides where you are weak.
Run it on a schedule
A single pass is a snapshot. Running the same set regularly turns it into a trend you can act on, so you can tell whether a content change moved the needle and spot a competitor pulling ahead before it becomes a pattern.
This is exactly what AI citation tracking automates, and the full approach is documented in the citation tracking methodology. To see where your site stands today, run the free AI visibility scan.