How metrics work
The shared measurement model behind every number in AgentMention — what a prompt, an answer and a mention are, and what each metric is calculated over.
Every metric in AgentMention is built from the same three things: a prompt you track, an answer an AI assistant gave to it, and the mentions and citations inside that answer. Understanding those three makes every number on the dashboard readable.
The unit of measurement
A prompt is a question a real customer might ask an assistant — "best running shoes for flat feet", "CRM for a 10-person agency". You organise prompts into groups and track them over time.
An answer is what one assistant replied to one prompt on one day. A prompt can produce several answers in a period — one per assistant, per run. The market the question is asked in is set by the prompt's group; see Geographies & languages.
A mention is your brand name appearing in the text of an answer. A citation is a link the assistant used as a source.
What the denominator is
This is the single most important thing to know about the numbers:
Metrics are calculated over prompts that received an answer, not over every prompt you track.
If a prompt produced no answer in the selected period — the assistant returned nothing, or the prompt was added after the last run — it is excluded from both the numerator and the denominator. A missing answer never counts as an absence of your brand, because that would silently punish you for a collection gap you did not cause.
The metric family
| Metric | Question it answers |
|---|---|
| Visibility | In how many answers does my brand appear at all? |
| Share of Voice | How much of the total brand talk is about me? |
| Position | When I do appear, how early am I named? |
| Sentiment | How favourably am I described? |
| Citations & Sources | Which pages do assistants read to build these answers? |
| Prompt gaps | Which questions do I lose entirely? |
Visibility and Share of Voice are deliberately different: one counts answers, the other counts mentions. A brand named three times in one answer scores the same Visibility as a brand named once, but three times the Share of Voice.
Periods and comparison
Every dashboard metric is calculated over a selected period — 24 hours, 7 days or 30 days, or a custom range — and compared against the immediately preceding period of the same length. A change value only appears once there are at least two comparable periods of data; before that, the metric shows a value with no delta. See Reporting periods.
Scope
Metrics are always scoped to one brand, one prompt group and the assistants you have enabled. Switching any of those changes every number on the page — a Visibility of 40% on ChatGPT and 12% on Perplexity are both true statements about different scopes.
Updated 2026-08-03