This paper reviews the current development and application of AI-powered mental health chatbots, focusing on their potential to support individuals experiencing anxiety and depression. Through an in-depth literature review and comparison of nine available tools, Woebot and Wysa emerged as the most widely used and cited in the academic and clinical domains. Both platforms are built on cognitive behavioral therapy (CBT) principles and demonstrate evidence of reducing depressive symptoms in controlled trials. The paper examines their key technical features, evaluates their adherence to the mHcode framework, and analyzes both shared and unique design elements. Despite their promise, these systems face substantial technical, ethical, and social challenges—including algorithmic hallucinations, data privacy concerns, and limited cultural or linguistic inclusivity. The discussion also highlights emerging issues such as the generation of hazardous content and sycophantic behavior due to current training paradigms. Future directions emphasize the need for improved factual accuracy, stronger content moderation, multilingual capabilities, and increased human oversight. While AI chatbots show potential as accessible mental health tools, their limitations must be addressed before they can be considered reliable substitutes for professional psychological care.