Public AI visibility benchmark · beta
The SoundingScore Index
SoundingScore combines favorable answer-text brand presence and deterministic citation authority across a shared, repeated buyer-intent benchmark. Every published score includes its confidence interval, sample, market, model panel and methodology version.
Audit your brandCurrent publication register
Current cohorts and verified rankings are available through the public data register. An empty rankings array means no result is currently published for the requested cohort; it must not be interpreted as a zero score.
Read the live public register (JSON) or open its machine-readable dataset description.
The register lists available cohorts. Retrieve ranking rows with one cohort's exact category, locale, market and period: https://bhsrltwdimueobppuylv.supabase.co/functions/v1/get-public-soundingscores?category={category}&locale={locale}&market={market}&period={period}.
How the beta score works
SoundingScore equals 100 multiplied by 75% favorable brand presence plus 25% verified citation authority. Positive mentions contribute fully, neutral mentions contribute half, and a citation counts only when a returned source URL matches the measured domain.
A normal GetSounding audit is a directional Sounding Snapshot. It does not enter this ranking. Public results require a shared category panel, repeated formulations and runs, fixed model coverage, validation thresholds, confidence intervals and complete evidence lineage. Read the SoundingScore method.
To qualify, at least 95% of planned responses must be valid overall, and those responses must represent at least 95% of the benchmark's pre-set weight. Every fixed model member must reach at least 90% on both measures.
The beta matrix uses 50 as the boundary for both mention coverage and perception quality. At least 50 is high; below 50 is low. A quadrant is assigned only when both 95% confidence intervals stay entirely on one side of their boundary. If either interval crosses 50, the result is Borderline.