This survey looks at the latest progress in detecting AI-generated voices, often called deepfakes. As voice synthesis technology becomes more advanced, it raises significant concerns about misinformation, fraud, and various malicious attacks, including identity theft and social engineering. These threats make it essential to have strong detection methods to protect individuals and organizations. In this paper, we explore different classifiers, including traditional machine learning techniques and modern deep learning methods, to assess their effectiveness in identifying deepfake audio. We summarize findings from various studies, explain the methodologies used, and compare how well different classifiers perform across a range of datasets. Our aim is to highlight the need for dependable detection tools and to point out important challenges and future directions for research in deepfake detection.