Time2Stop is an adaptive and explainable just-in-time adaptive intervention (JITAI) system designed to reduce smartphone overuse. It leverages machine learning to identify optimal intervention timings, provides transparent AI explanations, and collects user feedback to establish a human-AI loop for continuous model adaptation. The system includes a smartphone-based sensing app to collect user context and behavior, a cloud-based ML pipeline for feature extraction and prediction, an interface for delivering interventions and collecting feedback, and a human-AI feedback loop to update the model. An 8-week field experiment with 71 participants evaluated the system's effectiveness, showing that adaptive models significantly outperformed baseline methods in intervention accuracy and receptivity. Incorporating explanations further enhanced effectiveness, reducing app visit frequency by 7.0–8.9%. The system's adaptive and explainable features improved user trust and intervention effectiveness. Time2Stop demonstrates the potential of AI-based JITAI systems with human-AI loops to address smartphone overuse and other applications. The system's design integrates user feedback and explanations to enhance model adaptability and user engagement. The study highlights the importance of adaptive and explainable interventions in AI-based systems for improving user experience and trust.Time2Stop is an adaptive and explainable just-in-time adaptive intervention (JITAI) system designed to reduce smartphone overuse. It leverages machine learning to identify optimal intervention timings, provides transparent AI explanations, and collects user feedback to establish a human-AI loop for continuous model adaptation. The system includes a smartphone-based sensing app to collect user context and behavior, a cloud-based ML pipeline for feature extraction and prediction, an interface for delivering interventions and collecting feedback, and a human-AI feedback loop to update the model. An 8-week field experiment with 71 participants evaluated the system's effectiveness, showing that adaptive models significantly outperformed baseline methods in intervention accuracy and receptivity. Incorporating explanations further enhanced effectiveness, reducing app visit frequency by 7.0–8.9%. The system's adaptive and explainable features improved user trust and intervention effectiveness. Time2Stop demonstrates the potential of AI-based JITAI systems with human-AI loops to address smartphone overuse and other applications. The system's design integrates user feedback and explanations to enhance model adaptability and user engagement. The study highlights the importance of adaptive and explainable interventions in AI-based systems for improving user experience and trust.