Habitat thresholds are usually defined as “points of abrupt change” in the species–habitat relationships. Habitat thresholds can be a key tool for understanding species requirements, and provide an objective definition of conservation targets, by identifying when habitat loss leads to a rapid loss of species, and the minimum amount of habitat necessary for species persistence. However, a large variety of statistical methods have been used to analyse them. In this context, we reviewed these methods and, using simulated data sets, we tested the main models to compare their performance on the identification of thresholds. We show that researchers use very different analytical tools, corresponding to different operational definitions of habitat thresholds, which can considerably affect their detection. Piecewise regression and generalized additive models allow both the distinction between linear and nonlinear dynamics, and the correct identification of break point position. In contrast, other methods such as logistic regression fail because they may incorrectly detect thresholds in gradual patterns, or they may over or underestimate the threshold position. In conservation or habitat modelling, it is important to focus efforts efficiently and the inappropriate choice of statistical methods may have detrimental consequences.