The paper describes a systematic approach to the model reduction of large dimension fluid-structure-flight models, and the subsequent flight control design of very flexible aircraft. System nonlinearities may be due to the large wing deformations, the coupling between flexible and rigid body dynamics and/or flow separation at large angles of incidence. A nonlinear reduced order model is used to reduce the computational cost and dimension of the large-order nonlinear system for a practical control law design. The approach uses information on the eigenspectrum of the coupled system Jacobian matrix and projects the system through a series expansion onto a small basis of eigenvectors representative of the full-order dynamics. For a pitch-plunge aerofoil with structural nonlinearities, a controller based on reduced models was designed to alleviate gust loads. The approach to model reduction was also demonstrated for a two-dimensional problem with aerodynamics modelled using the computational fluid dynamics equations, and a flexible wing modelled using the geometrically-exact nonlinear beam equations. In all cases, the model reduction was found adequate to predict the large order system dynamics at a neglegible cost compared to that incurred by solving the nonlinear full-order system.