Water scarcity and rising global food demand pose critical challenges that call for more efficient and adaptive irrigation strategies. Traditional systems based on fixed schedules often fail to respond to dynamic soil and climate conditions, resulting in water waste and reduced productivity. Advances in Artificial Intelligence (AI) and Internet of Things (IoT) technologies now enable real-time, data-driven irrigation management. This paper presents a systematic review (2018–2023) of AI-based approaches—particularly time-series models such as Long Short-Term Memory (LSTM) networks, Artificial Neural Networks (ANN), and Support Vector Machines (SVM)—applied to irrigation optimization. A structured search across IEEE Xplore, ScienceDirect, Scopus, Web of Science, and Google Scholar identified over fifty relevant studies, of which thirty-four were critically analyzed following the PRISMA 2020 guidelines. The findings reveal that AI–IoT integration can reduce water consumption by 25–40% and increase crop yields by 5–12%. LSTM and hybrid models (e.g., CNN–LSTM, RF–LSTM) demonstrate superior performance in capturing temporal dependencies and improving irrigation scheduling accuracy. Despite these advances, challenges persist regarding data quality, computational cost, and scalability for smallholder farmers. The review concludes by outlining future directions centered on explainable AI, IoT–satellite data fusion, and low-cost edge computing to enable sustainable, scalable, and transparent smart irrigation systems.