Rank-based tests for detecting the presence of change point in a linear regression model are presented. We build upon Theil’s idea, namely to consider the slope of all connecting lines, instead of the original points. We, however, use the idea to construct change point rank tests in regression in a way that it reduces the problem to a two-sample problem. Wilcoxon type tests are proposed, with ranks assigned to the pooled sample, whose exact distribution is known under the null hypothesis of no change point. Simple linear regression models are considered, for both fixed and random design. Extensions to multivariate and errors-in-variables models are discussed.