You cannot improve what you cannot measure, and AI answers feel unmeasurable: private, probabilistic, different every session. They are not. Here is how to measure AI visibility with metrics rigorous enough to run a program on.
Mention rate: the percentage of sampled answers that name you, per question, per assistant. Share of voice: your mentions relative to each competitor on the same questions. Sentiment and framing: whether descriptions are accurate and favorable, scored consistently. Citation share: how often the sources behind answers are yours or carry you. Together they form a dashboard that behaves like any other growth metric.
Because assistants vary answers between sessions, a single check is noise. Reliable measurement runs each question multiple times per assistant per period, in clean sessions, and reports rates with trend lines. Weekly sampling across ChatGPT, Gemini, Claude, and Perplexity is the practical minimum; continuous is better.
Measure the questions that carry revenue: real buyer phrasings gathered from sales calls, support tickets, and community threads, anchored to the moments that trigger purchase. Twenty well-chosen questions beat two hundred generic ones. This moment-anchored approach is the foundation of Aethon’s Contextual AI Presence Mapping©.
For executives, roll the metrics into three numbers: overall mention rate (are we in the conversation), competitive share of voice (are we winning it), and accuracy score (is what AI says true). Add before-and-after answer screenshots for the work shipped each month. No marketing report lands harder than “here is what ChatGPT said in March, and here is what it says now.”
Tag inbound leads with “how did you hear about us” options that include AI assistants, and watch branded search and direct traffic alongside mention-rate gains. Brands consistently see AI-sourced buyers arrive better-qualified: the assistant already matched them to you. Aethon’s reporting ties visibility movement to these downstream signals.
By repeatedly running a fixed set of real buyer questions across ChatGPT, Gemini, Claude, and Perplexity and scoring mention rate, share of voice, sentiment, and citation share over time.
It varies by category maturity, but the benchmark that matters is trend and competitive share: growing rates and shrinking competitor gaps on the questions that drive revenue.
Weekly sampling minimum; continuous monitoring catches model-update shifts within days. Quarterly checks miss most movement.
Overall mention rate, share of voice versus named competitors, and accuracy of descriptions, plus concrete before-and-after answers showing the program working.
A manual baseline is a fine start: your top twenty questions, four assistants, recorded systematically. Sustaining reliable measurement at sample sizes that mean something is what platforms like Aethon are for.
The AI search visibility metrics KPIs worth reporting form a short stack. Coverage: of your category’s buying questions and moments, what share produce any answer that includes you. Share of voice: how often you are recommended versus each competitor, per assistant, the metric defined in AI share of voice. Position and framing: first recommendation or afterthought, accurate description or stale one. Source health: how many of the pages assistants cite about you are current. And revenue linkage: AI referred sessions, conversions and pipeline, which is the KPI that keeps the program funded.
Report them per assistant, ChatGPT, Gemini, Claude and Perplexity, and monthly, because model and source updates move answers faster than search rankings ever moved. The full metric definitions live in AI search visibility metrics.