Depression is a serious mental health disorder that impacts a significant number of people around the world. A major impact of depression is the persistent sense of sadness, loss of interest, and a decreased motivation. A crucial aspect of effective treatment is detecting and addressing depression at an early stage. As individuals often express their thoughts and emotions online, social media platforms have become valuable resources for mental health research. Textual data from social media can be analyzed to detect early signs of depression, providing an opportunity for timely intervention and support. In text analysis, machine learning algorithms are commonly used to detect depression based on predefined linguistic and emotional features. For traditional techniques, extensive feature engineering is required, and nuanced, context-specific language patterns associated with mental health conditions may be missed. By automatically learning these features, deep learning approaches identify complex patterns. Transformer models are capable of capturing subtle, context-rich markers of depression better than both machine learning and deep learning methods. This study provides an overview of the state-of-the-art machine learning, deep learning, and transformer-based models for detecting depression in text. Researchers and scholars conducting research on text mining and social media analysis will greatly benefit from this review.