To address the voltage violation issues caused by large-scale distributed photovoltaic (PV) integration in distribution networks, this paper proposes a comprehensive voltage sensitivity-based genetic algorithm (CVS-GA) for evaluating and optimizing distributed PV hosting capacity. First, by considering the nonlinear variation of voltage sensitivity, the traditional fixed-voltage-sensitivity calculation method is improved to reduce errors. A nonlinear programming model is then established to evaluate the PV hosting capacity, which accurately reflects the voltage limits under different PV installation schemes. This approach addresses the computational inefficiency and potential omission of boundary scenarios in traditional stochastic scenario simulation methods16. Second, a comprehensive voltage sensitivity calculation method is proposed, incorporating multi-timescale variations in PV and load power, as well as differentiated voltage regulation requirements across nodes. This method enhances the efficiency of genetic algorithms in optimizing PV capacity allocation through local iterative optimization. Finally, the effectiveness of the proposed algorithm is validated using the IEEE 33-node test system.