Depression is a pervasive global mental health challenge, often leading to severe outcomes such as suicidal intent if not identified early. This study introduces an AI-driven application that integrates facial emotion recognition using a CNN trained on the FERPLUS dataset with a generative AI chatbot to provide personalized, non-invasive support. The system combines facial analysis, personality profiling through Big Five test, and sentiment analysis of user input to create a holistic understanding of a user’s emotional and psychological state. The facial emotion recognition model achieved a validation accuracy of 80.98%, while the chatbot model showed significant improvement after fine-tuning its test accuracy increased from 40.25 to 97.91%, and F1-score rose from 0.5612 to 0.9732 demonstrating enhanced classification reliability and contextual response capability. Sentiment scores and personality traits especially high neuroticism served as important indicators for detecting emotional distress. Real-time alerts and actionable insights delivered via Telegram facilitate timely intervention. This multimodal, scalable system highlights the potential of AI in accessible mental health care.