Digital nudging embeds the concept of “nudging” from behavioral economics into digital governance contexts. By leveraging algorithmic recommendations, interactive interfaces, and real-time data feedback, it subtly guides public behavior without depriving individuals of their freedom of choice, and has gradually become an important new form of behavioral governance in the era of digital governance. However, existing studies have mainly examined digital nudging from the perspectives of behavioral economics or information systems, while lacking a systematic explanation of its embedded logic, theoretical foundations, and operational mechanisms within public administration. Against this background, this study reviews the evolutionary trajectory from traditional nudging to digital nudging and constructs a theoretical framework for digital nudging from three perspectives: behavioral science, public governance, and digital governance. This paper proposes a three-stage framework of “analysis-design-evaluation” for digital nudging in public administration. Furthermore, from the dual perspectives of “tool types-behavioral mechanisms”, this study systematically classifies digital nudging into five typical categories: scarcity-based nudges, default nudges, social nudges, reinforcement nudges, and feedback nudges. Additionally, it reveals the mechanisms through which these nudges influence public behavior, including choice architecture optimization, social norm information, personalized recommendations, reminders and notifications, interface design, and feedback reinforcement. While digital nudging can enhance governance precision and policy implementation efficiency, it may generate legal and ethical risks, such as weakened autonomy, cognitive manipulation, privacy leakage, algorithmic bias, and procedural unfairness. Finally, this study outlines future research agendas concerning the technological evolution, disciplinary embedding, institutional regulation, and legal-ethical boundaries of digital nudging.