Phishing is a prevalent social engineering attack that deceives users into revealing personal or sensitive data. The application of artificial intelligence (AI) techniques in detecting phishing attacks are explored in this literature review, highlighting machine learning (ML) and deep learning (DL) algorithms. The study aims to consolidate existing research, identify effective AI algorithms, and assess their performance. The study was conducted as a survey, with data gathered from IEEE Xplore, SpringerLink, MDPI, and ScienceDirect using PRISMA guidelines. The review included 21 peer-reviewed articles published between 2019 and 2024. The results show that AI-based methods, particularly hybrid models and convolutional neural networks (CNNs), have high accuracy in phishing detection, with XG Boost achieving the highest accuracy at 99.89%, followed by PILFER with 99.5%. Random Forest (RF) showed consistent performance across multiple studies. The review sheds light on the importance of diverse feature sets and numerous classifiers to enhance detection accuracy. However, challenges such as the evolving nature of phishing attacks and imbalanced datasets persist. The study recommends continuous updates to detection models and integrating multiple data sources for improved efficacy. Future research could usefully explore optimized learning parameters and hybrid approaches to enhance detection accuracy and scalability.