4 · Documentation
Presence and position
How presence, share of voice, place and range presence are calculated, and why only certain questions count.
Four figures describe where a brand stands in the AI answers of a market. They are calculated from the same stored answers, but not from the same questions.
Presence
Presence is the share of answers in which the brand appears. It is counted across the counting questions: questions with no brand in the wording and no segment. If the brand is already in the question, the mention would be a given. Segment questions would make the result depend on which segments happen to be measured.
The overview shows two values that may differ: presence in the latest run and presence across all runs. The second is the reliable one; the first shows the direction.
A mention counts when the judge has backed it with a passage that is found character by character in the answer. Without evidence the brand counts as not mentioned. See Evidence.
Share of voice
Share of voice is the brand's share of all brand mentions and sums to 100 % across all names. Per answer every brand counts at most once; otherwise whoever is repeated most often in the same text would lead.
The median of the ranking is shown next to it in the competition view, so that "21 %" has a reference.
Position
The place comes exclusively from ranking questions such as "What are the best …?" or "Name the five best …". A mere enumeration does not order; reading a position from it would not be reliable. The average place is the mean list position across all ranking questions in which the brand appears.
Two things deliberately stay empty rather than estimated:
- Brands without a ranking question show a dash, not a zero.
- Only brands entered in the project get a place. Names that were merely collected from answers get no rank, even with high presence. They can be adopted as competitors with one click; the matrix then extends retroactively across all runs.
A name enters the ranking from two mentions in counting answers. A name that appears in exactly one answer is usually not a competitor but chance; in the rice market that applied to 93 of 188 collected names.
Range presence and segments
A brand can be well known and still be missing from one product line. The good brand value hides that. Questions that name a product segment are therefore assigned a segment and count only there, never in the overall value. Range presence is the presence across the segment questions; the overview shows it next to the brand value and reports the gap as a finding when it exceeds the noise band.
Two rules keep segment values honest:
- No segment in which your own brand does not compete. A brand without an organic line gets a guaranteed zero in an organic segment, which costs questions and says nothing.
- No segment with only one question. With a single question every segment value is chance. At least two questions per segment, one recommendation and one ranking question.
What does not enter presence
- Segment questions count in the segment.
- Questions that name a brand (comparison, strengths and weaknesses, perception) deliver topics and brand image.
- Need questions ("What does diesel cost in Hamburg today?") measure whether the AI takes on a concrete task and are evaluated separately.
- Sentiment does not enter presence or position. In AI answers about four out of five statements about brands are positive; a score that counts sentiment turns this uniformity into an apparent strength.
Which question is of which type and how to build a good question set is explained under Questions.
Competitors cost nothing
Every named brand is counted from the same answers. Another competitor in the project therefore costs no measurement; only questions cost. The exception is strengths-and-weaknesses questions: they name the brand in the question text, so every brand asked about that way costs a question of its own.
Updated: 2026-09-16