In this article, the synchronization problem of uncertain chaotic neural networks (CNNs) is discussed by using delayed impulsive control. Aiming at the disturbances of parametric uncertainties and flexible delays in impulses, we adopt the idea of average impulsive delay (AID) to deal with these delays as a whole and derive a synchronization criterion of uncertain CNNs based on the condition that uncertain parameters are norm‐bounded. To be exact, in our results, the delays in impulsive control input are no longer limited by some strong constraints and can exist flexibly between two continuous impulsive moments. The results also clearly show that the delays in impulses are conducive to the synchronization of uncertain CNNs. Finally, an example is simulated to examine the applicability of the theoretical results.