Also called: Wilson interval, measurement confidence
A confidence interval is the honest range around a measured AI-visibility score that accounts for sampling noise — GeoMagics reports a Wilson interval so a number is never presented as more certain than the data supports.
AI engines answer probabilistically: ask the same question ten times and you may get several different framings, and a brand may be named in some runs and not others. A single run is therefore not a measurement — it is one draw from a distribution. A confidence interval expresses how much that underlying rate could plausibly vary given how many times it was sampled.
GeoMagics uses the Wilson score interval, which behaves well for proportions even at small sample sizes, and publishes the interval alongside every visibility score. If the interval is wide, the honest reading is 'we are not yet sure' — and the system says so rather than dressing up noise as insight.
The practical consequence is disciplined: a change that moves a score by less than the width of the interval has not been shown to work. Treating within-noise movement as a win is the most common way AI-visibility 'measurement' misleads, and reporting the interval is how that trap is avoided.