AI analytics
Metrics reference
Every number in GEOX is extracted from real AI answers by the analysis pass. Here is exactly what each one means — and what it deliberately excludes.
The denominator, once and for all
Rates are always computed over analyzed answers. Empty or failed model replies are stored but never analyzed, so they don’t drag your rates down — and it’s why the raw count on Questions can exceed the analyzed count on Brand Exposure.
Mention rate
The share of analyzed answers that explicitly name your brand or one of its aliases. The analysis pass reads for a real reference — incidental substring matches don’t count. This is the headline GEO metric: “when AI answers this category’s questions, how often do we exist?”
Average position
When an answer is a ranked list (“the best tools are: 1. … 2. …”), your brand’s 1-based spot in it, averaged. Answers where you’re absent, or where the answer isn’t an ordered list, don’t affect the average — so read it together with mention rate: a great average position on a 10% mention rate means you rank well on the rare occasions you appear at all.
TOP1 / TOP3
How often you land 1st, or anywhere in the top 3, across every analyzed answer — not just the ones that mention you. Unlike average position, these can’t be flattered by scarcity, which makes them the honest summary of “are we a default recommendation?”
Sentiment
The tone of each mention on a −1.00 to +1.00 scale, shown as a signed average. “Widely loved but pricey” lands slightly positive; “a popular but dated option” slightly negative. If you aren’t mentioned, sentiment is treated as neutral (0) rather than missing.
Citations
Source URLs an answer explicitly references, pulled verbatim and aggregated by domain on the Citation Sources page. Some models rarely cite sources, so this can stay sparse without anything being wrong.
Mention share
Your brand’s mentions versus each tracked competitor’s across the same set of analyzed answers — your share of voice, head-to-head. A competitor is counted by the same rules you are (name or alias, real reference), so the comparison is symmetric.
Reading them together
- Low mention rate, good position — you’re credible but invisible: work Citation Sources and publish more citable content.
- High mention rate, poor position — you’re in the conversation but not a default: close competitor gaps question by question.
- Good rate, negative sentiment — AI knows you but repeats criticism: check which sources the engines cite and what they say about you.
- Few analyzed responses — treat every other number as noise until more runs land.