In this paper, we investigate the real roots of a special class of square matrices, leveraging the properties of involutory and idempotent matrices. We focus on determining real roots for real orthogonal and symmetric matrices, demonstrating how involutory matrices facilitate this process. Our results show that a real involutory matrix of order $n$ with a positive determinant always admits a real root. Furthermore, for real symmetric matrices, we establish that a real root exists if every negative eigenvalue appears with even multiplicity. We also explore the structure of idempotent matrices, presenting a general block form derived through similarity transformations. Specifically, we prove that for invertible submatrices $A$ and $D$, along with arbitrary block matrices $B$ and $C$, a constructed matrix $P$ exhibits idempotency. An illustrative example is provided to clarify this construction, highlighting its application in generating idempotent and involutory matrices from simpler components. Additionally, we examine the root-Approximability of orthogonal matrices, showing that certain sequences of matrices converge to the identity while their powers approximate the original matrix. This work extends existing results on matrix functions and diagonalization, offering practical insights into the analysis and computation of matrix roots. Our findings contribute to the broader understanding of matrix theory, with potential applications in numerical linear algebra and functional analysis.