This paper deals with a frequency shaping, which means an improvement of frequency response, for fuzzy control systems with unknown non-linear plants by a learning method of neural network. We derive an important theorem, which is related to phase-lead compensation, by introducing concept of frequency characteristics such as gain crossover frequency and phase margin. Using the theorem, a fuzzy phase-lead compensator, which have parameters for effectively compensating phase characteristics of control systems, is constructed. After proposing a design procedure of the fuzzy phase-lead compensator for linear plants, we extend it to a design procedure for unknown non-linear plants. A frequency shaping for unknown non-linear plants is realized by Widrow-Hoff learning rule which is a basic learning algorithm of neural networks. Simulation results show the effect of frequency shaping by Widrow-Hoff learning rule.