Methods for analyzing the completeness of earthquake historical catalogs are applied to estimate which part of a historical catalog likely includes all earthquakes of a certain magnitude class after a certain time in a given area. Current methods of catalog completeness estimation are not sufficient in case of long‐term nonstationary seismicity data. Such methods typically require model presumptions, whereas the method described in this study does not. We develop a statistical approach based on a deviation criterion, which is applicable to stationary, nonstationary, and even multimodal nonstationary seismicity rates. The significance of the method is tested using synthetic catalogs extracted from stationary (exponential based on a Poisson process) and nonstationary (mixed bimodal Weibull–Weibull, gamma, and log‐normal) distributions. The method is applied to different parts of the Euro‐Mediterranean region based on long‐term seismicity data.