A surrogate-based optimization framework is proposed to exploit a reduced order model (ROM) as surrogate evaluator in aerodynamic design based on computational fluid dynamics (CFD) methods. The model is based on the Proper Orthogonal Decomposition (POD) of an ensemble of CFD solutions. Full POD and zonal POD models performances are analysed with respect to their suitability to find the global optimum in an evolutionary optimization frame. Indeed, reduced order models are used as fitness evaluator to improve the aerodynamic performances of a two-dimensional airfoil. Finally, the performances of various surrogate-based shape optimization (SBSO) methods are compared to the efficiency of data-fit assisted optimization and to the accuracy of a plain optimization, where, instead, each aerodynamic evaluation is performed with the high-fidelity model.