Depression is one of the most common mental health disorders that afflict people and productivity in society. This paper explores automatic depression detection using user-generated content on social media. Using machine learning and deep learning models, we thus design an efficient classifier system trained on a balanced set of depressive and non-depressive social media posts. We put together in our study techniques that have been learned from the SVM, random forest classifier, multinomial naïve Bayes, and long short-term memory network as well as the advanced methods of word embeddings such as Word2Vec and BERT. The models are classifying with accuracy ranging between 70% and 81.79%, offering scalability for the early stages of identification for depression. This result brings up the potential of AI to help advance diagnostics in mental health.