Online mental health surveys have long been challenged by respondent carelessness, a problem that has persisted for over a decade without a comprehensive solution. To tackle these issues, we designed and implemented the Digital Human-Based Interactive Mental Health Assessment System (DHMHAS) with college students as the pilot group, and employed mixed methods to validate its scientific credibility and interaction experience to comprehensively explore whether the system may retain scientific validity while increasing motivation to answer questions. We successfully confirmed that DHMHAS possesses robust psychometric properties. In particular, the system performed better than traditional methods in terms of criterion validity. Furthermore, DHMHAS enhanced the college students’ interactive experience by providing a high level of social presence and system usability, reducing cognitive load, and eliciting high levels of response differentiation in answers. They also offered overwhelmingly positive feedback on the system, recognizing its benefits for personal growth and development. Compared to traditional methods, DHMHAS offered significant improvements in delivering positive emotional experiences, deepening cognitive understanding, promoting immersive engagement, and increasing the accuracy of responses. This research lays a scientific foundation for developing innovative assessment tools and offers directions for optimizing future AI-based mental health evaluation methods.