Key points: The neuromotor system generates flexible motor patterns that can adapt to changes in our bodies or environment and also take advantage of assistance provided by the environment.We ask how energy minimization influences adaptive learning during human locomotion to improve economy when walking on a split‐belt treadmill. We use a model‐based approach to predict how people should adjust their walking pattern to take advantage of the assistance provided by the treadmill, and we validate these predictions empirically.We show that adaptation to a split‐belt treadmill can be explained as a process by which people reduce step length asymmetry to take advantage of the work performed by the treadmill to reduce metabolic cost.Our results also have implications for the evaluation of devices designed to reduce effort during walking, as locomotor adaptation may serve as a model approach to understand how people learn to take advantage of external assistance. In everyday tasks such as walking and running, we often exploit the work performed by external sources to reduce effort. Recent research has focused on designing assistive devices capable of performing mechanical work to reduce the work performed by muscles and improve walking function. The success of these devices relies on the user learning to take advantage of this external assistance. Although adaptation is central to this process, the study of adaptation is often done using approaches that seem to have little in common with the use of external assistance. We show in 16 young, healthy participants that a common approach for studying adaptation, split‐belt treadmill walking, can be understood from a perspective in which people learn to take advantage of mechanical work performed by the treadmill. Initially, during split‐belt walking, people step further forward on the slow belt than the fast belt which we measure as a negative step length asymmetry, but this asymmetry is reduced with practice. We demonstrate that reductions in asymmetry allow people to extract positive work from the treadmill, reduce the positive work performed by the legs, and reduce metabolic cost. We also show that walking with positive step length asymmetries, defined by longer steps on the fast belt, minimizes metabolic cost, and people choose this pattern after guided experience of a wide range of asymmetries. Our results suggest that split‐belt adaptation can be interpreted as a process by which people learn to take advantage of mechanical work performed by an external device to improve economy. Key points: The neuromotor system generates flexible motor patterns that can adapt to changes in our bodies or environment and also take advantage of assistance provided by the environment.We ask how energy minimization influences adaptive learning during human locomotion to improve economy when walking on a split‐belt treadmill. We use a model‐based approach to predict how people should adjust their walking pattern to take advantage of the assistance provided by the treadmill, and we validate these predictions empirically.We show that adaptation to a split‐belt treadmill can be explained as a process by which people reduce step length asymmetry to take advantage of the work performed by the treadmill to reduce metabolic cost.Our results also have implications for the evaluation of devices designed to reduce effort during walking, as locomotor adaptation may serve as a model approach to understand how people learn to take advantage of external assistance. [ABSTRACT FROM AUTHOR]