How AI Share of Voice Is Calculated

The canonical formula

Share of Voice = target valid mentions / (target valid mentions + configured competitor valid mentions) × 100.

This is a competitor-relative share, not market share, search volume or an overall quality score. The comparison universe, engine scope, question scope and dated measurement window must be known before the percentage can be interpreted.

The 4-step measurement process

Step 1: Use a defined comparison universe

Canonical SOV requires the target brand plus configured competitors. Without that comparison universe, SOV is not measured rather than guessed from the target brand alone.

Step 2: Collect compatible dated answer evidence

SOV Tracker records provider outcome and returned answer evidence for the tracked question scope. Technical/provider failure stays separate from a successful answer where no compared brand is mentioned.

Step 3: Count valid mentions

The analyzer counts valid occurrences of the target brand and configured competitors in the returned evidence using the canonical mention checks.

Step 4: Calculate competitor-relative Share of Voice

Raw SOV = target valid mentions ÷ (target valid mentions + configured competitor valid mentions) × 100. Prominence, sentiment, engine market share, question popularity and question volume are not added to this percentage.

SOV is not the whole visibility story

A brand can have the same SOV while appearing earlier or later in answers, being recommended more or less often, receiving different citation evidence, or being described with different sentiment. SOV Tracker therefore reports these as separate metrics instead of blending them into one /100 composite metric.

Discovery Visibility: found / answered on non-branded discovery questions.
Recommendation Rate: defined in the methodology but not measured yet; reports show it as not measured, never as 0.
Owned-Site Citation Coverage: observed citations to the target's own site on the eligible scope.
Sentiment: bounded positive / neutral / negative classification.
Prominence: where the first valid mention appears inside the returned answer; not a universal rank.
Brand Accuracy: whether the returned answer represents business-confirmed brand facts accurately on its eligible evidence scope.

40% Measured SOV Illustration: 2 valid target mentions and 3 valid competitor mentions produce 2 / 5 = 40%.

0% Measured zero Valid answers came back, a competitor set exists and competitors were named but the target brand never was: SOV is a measured 0%.

— Not measured If no brand in the set was named (0/0), SOV is not measured. No valid comparison universe or unavailable provider evidence means SOV is not measured. A technical failure is never rewritten as zero.

What does prominence mean?

Prominence describes where the first valid brand mention appears inside one returned answer, represented as an answer decile in the canonical analyzer. It is not a market rank, recommendation rank or provider-wide ranking, and it does not change raw SOV.

Engine and plan scope

The no-signup Free Scan uses ChatGPT with five automatically generated non-branded questions and is a limited Discovery scan, not competitor-relative SOV monitoring. Paid plans include ChatGPT + AI Overviews + Gemini + Perplexity + Claude; canonical SOV comes from the compatible synchronized full measurement window.

What the metric does not prove

It is not market share, revenue share or total search demand.
It is not a universal ranking across all AI users and questions.
A before/after movement does not by itself prove one specific change caused the movement.
A provider error is not evidence that the brand is invisible.

For the canonical current contract used by customer-facing surfaces, see the Methodology page.