["Economica, EarlyView. ", "\nAbstract\nRegular exercise offers well‐documented health benefits, and mobile applications are increasingly used to promote physical activity. In this study, we conduct a large‐scale randomized controlled trial with 20,187 users of the WeWard app in France to evaluate the effectiveness of different message framings and intervention durations on app engagement and walking behaviour. Participants were randomly assigned to receive one of three types of messages, each leveraging distinct behavioural features such as peer comparison, self‐comparison or a generic reminder, over a period of one or three weeks. Our results show that peer‐comparison messages positively influence the extensive margin of app utilization. However, despite these increases in app usage, neither type of message led to a significant increase in the number of steps walked. These findings highlight the limitations of relying on engagement metrics alone when evaluating app‐based interventions aimed at behaviour change.\n"]