As awareness of the importance of mental health continues to grow, there is an increasing demand for innovative diagnostic solutions that prioritize both accuracy and accessibility. This paper explores the development and validation of an AI-driven diagnostic tool that integrates facial expression analysis with textual data analytics to enhance the precision of mental health diagnoses. Utilizing the “Synthetic Therapy Conversations” dataset alongside over 10,000 facial expression datasets, we have developed a multimodal AI tool that synergistically combines textual content and facial expressions to assess emotional states, it simulates the human brain’s comprehensive ability in processing emotional and linguistic information. According to accuracy, the accuracy rate is higher when version-RFB is used for facial detection, ResNet-18 is used for facial expression recognition, and Random Forest is used for text analysis.