Main metrics
Orchestor measures how often brands appear, how they compare with competitors, their order of appearance, and how they are described in AI answers to tracked prompts. Start by reviewing these four metrics for the same period and conditions.
Four brand metrics
| Metric | What it measures | How to read it |
|---|---|---|
| Visibility | The share of answers that mention the brand | Higher means the brand appears in more answers |
| Share of voice | Your brand's share of mentions across comparison brands | Higher means more mentions relative to comparison brands |
| Average position | Where the brand appears in answers that mention it | Closer to 1 means an earlier appearance |
| Sentiment | How positively the brand is described | Displayed on a 0–100 scale, with 50 as the neutral reference |
These metrics describe the answers you track. They do not measure an AI service's total audience, search demand, or revenue share.
Count metrics
Check the underlying counts before interpreting a change in percentages.
| Metric | What is counted | How to read it |
|---|---|---|
| Mention count | Answers that mention a brand | Repeated mentions of the same brand within one answer count once |
| Answer count | Collected answers | Includes answers that do not mention the brand |
| Citation count | Recorded citation occurrences in answers | One answer can have several citations, so this differs from answer count and unique URL count |
For example, if 40 of 100 collected answers mention your brand and all 100 answers contain 250 recorded citations, your brand has 40 mentions, the answer count is 100, and the total citation count is 250.
Executions and answers
Attempted executions and collected answers are different counts. Failed or pending executions are not collected answers. Current brand analytics uses collected answers as its measurement base, rather than an execution count covering all attempts.
Also check each view's scope. Answer and mention counts in the prompts list aggregate the prompt's accumulated results. They may differ from counts in a report with a selected date range.
Citation metrics
| Metric | What it tells you | Details |
|---|---|---|
| Citation share | A domain or URL's proportion of all scoped citations | Counts, denominator, and example |
| Citation rank | Which sources lead when ordered by citation count or share | Sorting and reading the order |
Citation share uses citations as its denominator. Subtracting it from 100% does not give the percentage of answers without citations. Treat brand mentions, website citations, and information retrieval as separate observations.
Fields in the prompts table
These docs use Visibility, Share of Voice, Position, and Sentiment. The prompts table labels Share of Voice as SOV.
Alongside performance metrics, the table describes demand and properties of the prompt itself. Branding and consideration stage classify the question; they do not score the brand's performance.
| Column | Meaning | Details |
|---|---|---|
| Volume | Relative demand for a question or similar intent within an industry. The current list has no volume value and displays — | Set up prompts |
| Branding | Whether the question names your brand or a competitor: Branded or Non-branded | Organize with topics and tags |
| Consideration stage | Where a question falls in the funnel: Information gathering → Research & comparison → Purchase & sign-up. This is not an observed customer journey | Organize with topics and tags |
| Tags | Custom labels for organizing, filtering, and comparing prompts | Organize with topics and tags |
Mentions identifies brands appearing in the answers. Web search is the share of answers for which web search ran. A brand name in the question, a brand mention in the answer, and a web search are separate observations.
Different questions, even for the same answers
Suppose 100 answers include 40 mentions of your brand and a combined 60 mentions of comparison competitors.
- Visibility is 40%: divide the 40 answers mentioning your brand by all 100 answers.
- Share of voice is 40%: divide your 40 mentions by the 100 total mentions of your brand and competitors.
- Average position averages your brand's order of appearance in answers that mention it.
- Sentiment aggregates scores from answers where descriptions of your brand could be evaluated.
If competitor mentions increase to 160 while your brand stays at 40, visibility remains 40% and share of voice falls to 20%. Multiple brands can appear in one answer, so the sum of brand mentions can exceed the answer count.
Keep comparison conditions consistent
Compare the same periods, prompts, AI models, regions, and other conditions. Adding or removing prompts or comparison brands can change the denominators.
Check answer and mention counts alongside percentages. Two mentions in 4 answers and 50 in 100 both give 50% visibility, but represent different sample sizes. When combining periods, add the underlying counts and recalculate instead of taking a simple average of daily percentages.
Distinguish brand mentions from website citations
A brand appearing in an answer and its website being cited are different events. An answer may mention your brand without linking to your website, or cite your article without mentioning your brand.
Use these four metrics to understand brand presence, and Sources and citations to see which pages were used as evidence.
Check metric changes against the answers
Filter to the models or prompts that changed, then read the text in Answers. Review competitor comparisons, the features discussed, and cited URLs to decide which descriptions may need updating. A metric change alone does not prove that a particular action caused it.