The inherent conservatism in standard norm-based bounds for the sensitivity of the continuous-time algebraic Riccati equation is discussed and alternative sensitivity measures are introduced. These measures can be used to model a variety of situations where uncertainty in the data lead to an uncertain solution of the equation, and can be used to provide a more realistic evaluation of sensitivity than conventional bounds. The algorithmic sensitivity is discussed in this context. An efficient statistical approach for accurately estimating the sensitivity is described and examples of its different possible uses are given.