Understanding the process of human motion perception is essential for optimizing the motion cueing algorithms (MCAs) used by motion simulators. Previous research has shown that different motion cueing error types can be distinguished, at least some of which are rated differently by humans for equal error magnitude. Optimizing a motion simulation for both error magnitude and error type requires a method to objectively identify the presence of different motion cueing error types from simulation input and output traces. This paper presents a wavelet-based motion cueing error detection algorithm that exploits the time and frequency characteristics of these error types. A simulator experiment is presented through which parameters of the algorithm are determined. It is shown that motion cueing errors can be detected by the algorithm without prior knowledge of the MCA used. The need for the algorithm is demonstrated by showing that motion cueing error types can be divided in clusters of severity. Additionally, it is shown that in curve driving, given a reference of unity gain, the lateral specific force can be scaled down to 70% and up to at least 130% before subjects indicate motion to be too weak or too strong. A required input and output coherence of 0.7 is found for simulator input and output to be considered coherent. The motion cueing error detection algorithm can be used to improve motion cueing algorithms in the future.