How do Decision Trees behave with correlated features?

Decision Trees may arbitrarily select one feature among correlated ones, leading to unstable feature importance.

When features are highly correlated, decision trees tend to pick one feature for splitting and ignore the others, even if they carry similar information. Small changes in data can cause the tree to switch between correlated features, making the model unstable. This also leads to misleading feature importance scores, where importance is split unevenly among correlated variables.