Skip to main content
Definitions for each AI visibility metric Ryze reports, what it is calculated from, and how to read it.

​
Mention rate

The percentage of tracked answers in which your brand name appears in the answer text.
A mention does not require a link. A model can name a brand from training data without reading any page. Mention rate therefore reflects brand establishment more than content quality.

​
Citation share

The percentage of all source links across tracked answers that point to your domain.
Counted by domain rather than by answer, so one answer citing four of your pages does not overstate performance. This is the metric most responsive to content work. See citation share.

​
Share of voice

Your mentions as a proportion of all brand mentions across the tracked answer set, so all tracked brands sum to 100%. Answers the competitive question: when a buyer asks, how often is it us versus them. See share of voice in AI.

​
Answer coverage

The percentage of tracked prompts where your brand appeared at all — mentioned, cited, or both. The broadest presence measure. Useful as a top-line, too coarse for diagnosis.

​
Competitor overlap

Which competing brands appear in the same answers as you, and how often. Identifies your real competitive set in AI answers, which frequently differs from the one you assume — models often surface adjacent categories or established players you do not track commercially.

​
Position within the answer

Where your brand falls in the answer body — first recommendation, part of a list, or a passing mention at the end. First-position mentions carry disproportionate weight in a generated answer, because most readers stop early.

​
Visibility by assistant

The share of one assistant’s tracked answers that name a brand — the same measure as mention rate, but broken out per answer engine and shown for the top brands side by side.
Ryze shows this as the Visibility by Assistant matrix on the GEO tab: brands as rows, assistants as columns, your own brand highlighted. Read across a row to see where a brand is strong; read down a column to see which brands an assistant favours. Differences between columns are common — an assistant that leans on a particular source set can name brands the others ignore.

​
Reading these numbers

Compare against a baseline, not in isolation. A 4% citation share means nothing without knowing what it was last month and what the leader holds. Expect volatility. Only a minority of cited URLs repeat between two runs of the same prompt. Week-to-week movement is largely noise; monthly trend across many prompts is signal. Segment by query type. Aggregate numbers hide the useful structure. Platform-worded prompts behave completely differently from problem-worded ones and should be read separately. Mentions and citations move independently. Rising citations with flat mentions is normal and expected early — it means your content is being found before your brand is established.