Reference
Prompts metrics
How each card on the Prompts overview and prompt detail pages is calculated.
All Prompts metrics read the brand mentions recorded for each answer.
Filter by tags
Section titled “Filter by tags”Every card on the page follows the tag filter. A prompt is kept when it carries at least one of the selected tags; its answers are then the only ones counted.
Overview tiles
Section titled “Overview tiles”Answers
Section titled “Answers”The number of AI answers to your prompts created in the date range, after the tag and model filters. The subtitle is the number of prompts.
Models
Section titled “Models”The number of distinct AI models that produced those answers. Besides ChatGPT, Gemini and Claude, each active prompt also gets a Google AI Mode answer per scan; Google AI Mode counts as one model of its own, separate from Gemini, here, in the model filter and on every card that splits by model.
AI Share of Voice
Section titled “AI Share of Voice”Each brand’s slice of the pool of brand mentions:
share of voice = mentions of the brand / mentions of all brands × 100mention rate = mentions of the brand / answers × 100The slices of the chart sum to 100%. The mention rate, shown on hover, is the share of answers that name the brand.
Example: brand A is mentioned in 30 answers, B in 20, C in 10 → A 50.0%.
The tile with the same name on the Dashboard uses the mention rate of your own brand instead.
Average sentiment
Section titled “Average sentiment”The mean sentiment score over every answer in the date range, rounded to two decimals, then labelled:
| Score | Label |
|---|---|
| above 0.3 | Positive |
| −0.3 to 0.3 | Neutral |
| below −0.3 | Negative |
Example: three answers score 0.8, 0.5 and −0.1 → mean 0.40 → Positive.
Visibility
Section titled “Visibility”The share of answers in the date range that mention any of your own brands:
visibility = answers mentioning an own brand / answers × 100This measures how often you come up at all; share of voice measures how large your slice is once brands are named.
Rankings
Section titled “Rankings”Brands sorted by how many answers mention them.
| Column | Calculation |
|---|---|
| Visibility | mentions of the brand / mentions of all brands × 100; sums to 100% |
| Position | Average position, see below |
| Trend | Relative change in mentions over the last 7 days versus the 7 days before, independent of the date range |
| Sentiment | Mean sentiment score of the answers that mention the brand, shown as a coloured dot |
Position
Section titled “Position”Within one answer, the brands it names are ordered by where each brand name first appears in the text; the first brand is position 1. A brand’s position on the card is its average over the answers that name it. Lower means the model brings you up earlier.
Example: an answer names brand B before brand A → B = 1, A = 2. Over ten answers where A is named first eight times and second twice, A’s position is 1.2.
Visibility trend
Section titled “Visibility trend”Per day (per week for ranges longer than 92 days) and per brand:
line value = mentions of the brand that day / mentions of all brands that day × 100The lines sum to 100 on each day. A day without answers shows 0.
Top domains
Section titled “Top domains”The websites AI answers cite when answering your prompts. A domain counts once per answer, whether it appears in the answer’s sources or in the text.
percentage = answers citing the domain / answers × 100The ten most cited domains are shown.
Example: cited in 18 of 120 answers → 15%.
Recent answers
Section titled “Recent answers”The newest answers first, with a short preview, sentiment, mentioned brands and sources. Open an answer to read it in full.
Prompt detail
Section titled “Prompt detail”The prompt page repeats the Rankings and Visibility trend cards for one prompt, with the
prompt’s visibility score (answers mentioning an own brand / answers × 100, one decimal).
The score, the rankings and the trend count only the prompt’s answers within the selected
range, up to and including the end date of a custom range. Sentiment per brand on that page
is the exception: it is averaged over every answer the prompt has ever received.