Poisson regression is a type of generalized linear model (GLM) that is used to model count data. It assumes that the response variable follows a Poisson distribution, where the mean of the distribution is equal to its variance. This makes it suitable for modeling the number of events occurring within a fixed period or area.