Why do Decision Trees create axis-aligned decision boundaries?

Decision Trees create axis-aligned splits because each decision evaluates only one feature at a time.

Standard decision trees (CART, ID3 variants) evaluate splits using a single feature threshold, such as x_i <= t. This results in rectangular partitions of feature space, known as axis-aligned boundaries. While this simplifies optimization and improves interpretability, it limits the model's ability to represent diagonal or nonlinear boundaries efficiently, often requiring many splits to approximate complex decision surfaces.