Metrics

The three GEO metrics build on each other, from "are we there" to "are we the source".

Mention rateCopy Anchor Link

mention rate = runs mentioning your brand / total runs (in the window)

The floor metric: of all the answers your prompts generated, how many name you at all. It feeds the Growth Score's AI Visibility pillar with the largest weight (60%).

Reading it: a mention rate that holds steady while you publish more is common — mention rate moves when the share of answers that include you changes, not when you write more. Track it per engine: holding steady in Perplexity while dropping in ChatGPT points at OpenAI-side content, not your site.

Citation rateCopy Anchor Link

citation rate = runs citing your domain / total runs

Mentioned ≠ cited. An answer can name you while citing a competitor's page as the source. The citation analysis aggregates, over the measurement window, which domains the engines actually linked — the answer to "why is my competitor quoted instead of me":

• engines overwhelmingly cite pages that directly answer the question — comparison tables, definitions, step-by-step guides — not homepages,
• being cited is the mechanism by which GEO drives traffic: the citation is the link.

Closing a mention/citation gap is usually a content-structure job: make the answer to each tracked prompt exist as a clearly structured, crawlable section on your site. The Content Gap briefs give you the writing scaffold for the keyword side of that work — they are generated from content gaps, not from individual prompt answers.

Share of voice (SOV)Copy Anchor Link

SOV = your brand mentions / all brand mentions (you + competitors)

Zero-sum: your SOV rises exactly when a competitor's falls. SOV is computed against your tracked competitors — aliases included — so the list defines the game board.

SOV is the most strategic of the three: a mention rate can look healthy while a competitor quietly out-mentions you across every prompt. The benchmarks page visualizes exactly that.

Windows and data_sourceCopy Anchor Link

Metrics are computed over a lookback window (?days=, max 90): summary, SOV and the per-engine breakdown default to 14 days; the daily series defaults to 30 days. Every individual run carries its data_source label (live vs sampled) so you can tell them apart row by row in the run list — but the aggregate numbers (summary, SOV, the trend series, per-engine rollups) currently include all runs together, with no data_source filter parameter. When precision matters, eyeball the run list's labels before quoting an aggregate. See Data sources.