<p>Flood prediction can be challenging due to the complexity and non-linearity of its hazards, making traditional methods limited and less efficient. Thus, the interest in using machine learning and deep learning to overcome these challenges has increased over the years. This systematic review provides an overview of the current state of the flood prediction field using machine learning and deep learning models. It examines its evolution over the past two decades. It provides insights into the most commonly used models for various prediction tasks and timeframes. It also examines the most suitable input data for models adopted by researchers and explores the performance metrics employed. Our study synthesises 237 Scopus-indexed and English-language papers, retaining the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement as the primary reporting framework. The domain has experienced significant growth since 2021, making major contributions in the field of flood prediction. Long Short-Term Memory (LSTM), Artificial Neural Network (ANN), and Convolutional Neural Network (CNN) are the most commonly used models, demonstrating the superior performance of Deep Learning models over Machine Learning models in flood forecasting. Hybrid and ensemble models are found to be very beneficial in terms of prediction accuracy and efficiency, especially when dealing with complex spatiotemporal inputs. Discharge and flow forecasting is the primary prediction task among hydrologists, and rainfall and streamflow stand out as the primary predictor inputs. However, most studies focus on short-term-based prediction, leaving a gap in forecasting long-term floods that can inform large-scale water resources management strategies.</p>