Help centerReference
Metrics explained
Every number in GEO Radar: what it counts, how it is calculated and when it shows a dash.
Every number in GEO Radar is counted from real AI answers in the selected range and filters. This page lists how each one is calculated, how many answers it needs, and how to read the markers next to it.
How answers are read
After a scan, an AI reader goes through every answer once. For each brand the answer names, it records:
- Rank
- The order the answer presents brands in. 1 is named first.
- Role
- Recommended, compared, mentioned, or warned against.
- Sentiment
- 0 to 100, where 50 is neutral.
- Quote
- The exact words that show it. A quote only counts if it appears word for word in the answer.
- Praised / criticized
- Up to three strengths and three weaknesses the answer gives.
A tracked brand counts only when its name, an alias, its domain or a safe spelling variant appears in the answer. Brands you don’t track are saved too; they feed AI also names.
Brand metrics
- Visibility
- Calculation: Answers naming the brand ÷ all answers
- Shows a number after: 10 answers
- Recommended
- Calculation: Answers recommending the brand ÷ answers that recommend any product
- Shows a number after: 5 answers that recommend a product
- Position
- Calculation: Average rank when named, plus the share of mentions where it is named first
- Shows a number after: 3 mentions
- Sentiment
- Calculation: Average 0–100 across mentions. 60+ is Positive, 40 or less Negative
- Shows a number after: 3 mentions
- Share of voice
- Calculation: Each mention weighs 1 ÷ its rank. The brand's total weight ÷ the weight of all tracked brands
- Shows a number after: 10 answers
| Metric | Calculation | Shows a number after |
|---|---|---|
| Visibility | Answers naming the brand ÷ all answers | 10 answers |
| Recommended | Answers recommending the brand ÷ answers that recommend any product | 5 answers that recommend a product |
| Position | Average rank when named, plus the share of mentions where it is named first | 3 mentions |
| Sentiment | Average 0–100 across mentions. 60+ is Positive, 40 or less Negative | 3 mentions |
| Share of voice | Each mention weighs 1 ÷ its rank. The brand's total weight ÷ the weight of all tracked brands | 10 answers |
Below the minimum the value shows “—”, and charts leave a gap. A dash means “not enough answers to say”, not zero.
Source metrics
- Appearance
- Calculation: One source in one answer. Repeats in the same answer count once.
- Citation rate
- Calculation: Cited appearances ÷ all appearances
- Answer reach / In answers
- Calculation: Answers using the source ÷ all answers in range
- Gap answers
- Calculation: Answers where the source is cited, a competitor is named, and you are not
- Opportunity
- Calculation: Gap answers × log₂(1 + prompts) × (1 + cited share). High = top 20% of sources in view, Medium = next 40%, Low = the rest
| Metric | Calculation |
|---|---|
| Appearance | One source in one answer. Repeats in the same answer count once. |
| Citation rate | Cited appearances ÷ all appearances |
| Answer reach / In answers | Answers using the source ÷ all answers in range |
| Gap answers | Answers where the source is cited, a competitor is named, and you are not |
| Opportunity | Gap answers × log₂(1 + prompts) × (1 + cited share). High = top 20% of sources in view, Medium = next 40%, Low = the rest |
Changes, ranks and “New”
Markers
- ↑3.4Up 3.4 points: better
- ↓1.8Down 1.8 points: worse
- ↓0.6Position: a lower number is better
- NewAbsent in the previous period
- ↑Moved up a rank (tooltip: the old rank)
- ↓Moved down
Ties share a rank. After a model change, both periods use only the models that answered in both, and the tooltip says “same models: …”.
Most important first
Answers (Answer audit)
Each answer gets a score from the largest situation that applies, plus extras:
- A competitor is recommended and you are not named
- Points: 45
- You are named, but the answer is negative about you
- Points: 40
- A competitor is named and you are not
- Points: 30
- A competitor is ranked above you
- Points: 25
- The answer recommends tools without you
- Points: 15
- Extra: your site is cited but your brand isn't named
- Points: +15
- Extra: a competitor's own site is cited
- Points: +10
- Extra: you are missing on some of the models for this prompt
- Points: up to +15
| Situation | Points |
|---|---|
| A competitor is recommended and you are not named | 45 |
| You are named, but the answer is negative about you | 40 |
| A competitor is named and you are not | 30 |
| A competitor is ranked above you | 25 |
| The answer recommends tools without you | 15 |
| Extra: your site is cited but your brand isn't named | +15 |
| Extra: a competitor's own site is cited | +10 |
| Extra: you are missing on some of the models for this prompt | up to +15 |
Answers that recommend products or name competitors weigh most. Older answers count less, halving every 7 days. Needs attention holds answers with 25 points or more; Wins holds answers that recommend you in the top 3.
Prompts
A prompt ranks higher when its answers often recommend products, you are rarely named, and a competitor leads you. It gets a boost if your visibility dropped 5 points or more. It needs 10 answers.
Evidence lists
Answers behind a source, a gap or an action are listed in this order: a competitor recommended while you are missing, then a competitor named while you are missing, then answers where you are missing, then the rest. Within each group, cited comes before retrieved, and newer before older.