Arrow Define the unit and denominator before calculating AI share of voice. One approach divides a brand’s qualifying mentions by all qualifying mentions across a fixed competitor set. A different metric measures the percentage of valid answers mentioning the brand. These calculations answer different questions and should have different labels.
Write the counting rules before reviewing the results
Decide whether a brand counts once per answer or once per occurrence, how aliases are matched and which competitors are included. Distinguish a favorable recommendation from a passing or negative mention. A repeated brand name within one response should not create an accidental advantage.
Keep the same question panel and collection conditions when comparing periods. Changing the competitor set changes the denominator. Report that change clearly instead of interpreting every movement as improved visibility.
Show coverage alongside the percentage
Valid responses without the brand belong in a mention-rate denominator. Failed requests are not negative answers: show their count separately and explain the observed coverage. Do not combine incomparable engines or markets into one score without stating the weighting.
For a hypothetical set of ten valid answers, a brand mentioned in three has a 30% answer-level mention rate. If there are twelve qualifying brand appearances across the defined competitor set and three belong to that brand, its appearance share is 25%. Neither figure measures actual market share.
What to check
- Publish the formula, competitor set and counting unit.
- Keep a stable comparison panel or label breaks in the series.
- Show valid answers, non-mentions and collection failures separately.
A practical example
A report displays 30% answer-level mention rate and 25% share of counted competitor appearances, with the formulas beside them. The team avoids presenting the two values as interchangeable.
Sources and further reading
These steps are Arrow AI's implementation guidance. The example is illustrative; it is not a measured customer result. Publication and source access do not guarantee an AI recommendation.
Apply this to your website
Identify the source gaps and decide what to improve with Arrow's free audit. Review what Arrow GEO measures before choosing a reporting scope.