Today an average transmission line is loaded more heavily than ever before, and this has given rise to a serious problem of voltage instability. The authors present a novel method for power system voltage instability estimation and improvement using ANN. The method is based on the fact that reactive power injections at critical buses of the power system help to steer the system away from a developing voltage collapse. The location and quantum of reactive power support has been computed based on sensitivity. The sensitivity matrix is formulated by cascading the output/input sensitivity of multiplayer perceptron model with that of input features versus reactive power injections. The effectiveness of the proposed method is demonstrated on Ward-Hale 6-bus, IEEE 14-bus, and IEEE 30-bus test systems. The method can be effectively used to make the system secure against voltage collapse condition in system planning and online operation.