A researcher claims that for any dataset of $n$ non-negative numbers, the relationship between the arithmetic mean ($A$), geometric mean ($G$), and harmonic mean ($H$) is $G^2 = AH$. Is this statement universally true?