Extensive adoption of social media has created a rich source of user-generated data, enabling advanced AI-driven analysis of psychological behaviour. This learning investigates the application of machine learning (ML) and natural language processing (NLP) techniques to predict psychological traits and mental health conditions from social media posts. Using supervised learning models counting Logistic Regression, Support Vector Machines (SVM), and Random Forest classifiers—we demonstrate the effectiveness of text classification in identifying behavioural patterns with high accuracy. Our findings highlight the potential of AI in early mental health risk detection, offering scalable and non-invasive insights for psychological assessment. However, deploying such systems requires careful consideration of ethical challenges, with data privacy, algorithmic bias, and interpretability. The study underscores the need for further research to enhance model generalizability, incorporate multimodal data (e.g. images, temporal activity), and establish ethical frameworks for responsible AI deployment in mental health analytics.