The formula
Share of Voice = brand mentions / (brand mentions + configured competitor mentions) Γ 100.
This is a share, so the product needs a real comparison universe. If competitors are not configured, or there are no valid comparison mentions, SOV is not measured. Missing measurement is not converted to zero and is not converted to one hundred.
The 4-step measurement process
Step 1: Collect a dated answer
SOV Tracker runs the tracked prompt on the engines included in the workspace plan and records the provider outcome and returned answer evidence for that scan.
Step 2: Count valid mentions
The analyzer counts valid target-brand mentions and configured competitor mentions. Provider failure is kept separate from a real negative result.
Step 3: Calculate raw Share of Voice
Raw SOV = brand mentions Γ· (brand mentions + configured competitor mentions) Γ 100. It is a percentage share. Position/prominence is not added to this percentage.
Step 4: Calculate Visibility Score separately
Visibility Score is a separate 0β100 product score for answer visibility/prominence. Keeping it separate prevents a position-weighted score from being mistaken for Share of Voice.
Why SOV and Visibility Score are separate
A percentage share answers 'what share of valid mentions belongs to my brand inside this comparison set?' A visibility score answers a different question: 'how visible/prominent is the brand in the returned answer evidence?' Combining them makes the percentage impossible to interpret correctly.
What does position mean?
The analyzer's first-mention position is a coarse response decile (where in the returned answer the first valid mention appears), not a universal list ranking. It may contribute to Visibility Score, but it does not change raw SOV.
Engine and plan scope
The no-signup Free Scan and registered Free plan use ChatGPT. Paid plans include ChatGPT + AI Overviews + Gemini + Perplexity + Claude. Results should always be interpreted together with the engines and prompts actually measured in that dated scan.
What the metric does not prove
- It is not market share or revenue share.
- It is not a universal ranking across all users and prompts.
- 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.