In the study of exceptional point (EP)-based sensors, the concrete form of the output spectrum is often dismissed, and it is assumed that there is a corresponding relation between the peaks/valleys in the transmission spectrum and the real parts of the eigenvalues of the system. We point out that this assumption does not always hold. An effect, which is mathematically similar to electromagnetically induced transparency (EIT), may result in a ‘pseudo spectrum splitting’ that does not correspond to the splitting between the eigenvalues. The effect shall be taken care of when designing an EP-based sensor since it may cause measurement error and misunderstanding such as recognization of the spectrum splitting as the eigenvalue splitting at the exceptional point. We also propose to intentionally utilize this ‘pseudo splitting’ to design a sensor, which does not work at an EP, that has an EP-like spectrum splitting.