Flying Ad Hoc Networks (FANETs), which are made up of Unmanned Aerial Vehicles (UAVs), have changed the way aerial communication systems work by letting them work together automatically in mission-critical situations like crisis response, military surveillance, and environmental monitoring. However, FANETs’ high node mobility, changeable architecture, limited energy resources, and narrow communication ranges make standard routing algorithms difficult. The intelligent routing protocols created for UAV-based FANETs are thoroughly reviewed in this study, with a particular emphasis on the proactive, reactive, and hybrid routing paradigms. This article critically examines how Reinforcement Learning (RL) can be included into routing algorithms to improve flexibility, scalability, and decision-making in unpredictable aerial settings, going beyond conventional approaches. RL-based routing techniques, such as Single-Agent, and Multi-Agent Reinforcement Learning (MARL), improve latency, energy-efficient path selection, connection stability under high mobility, and packet delivery ratio (PDR). This paper offers essential insights for researchers and practitioners seeking to develop advanced routing protocols for UAV-based networks.