Multicast routing refers to the transmission of packets to a group of nodes identified by a single multicast group address. It plays a critical role in supporting applications that require group communication such as video conferencing and file distribution. One particularly challenging environment for multicast routing is Mobile Ad-hoc Networks (MANETs). The major problems facing routing in such networks are node mobility, frequently changing topology, unstable wireless links, and limited transmission range. Despite these challenges, several multicast applications like data base initialization and file distribution applications require reliable and efficient delivery of data. Various approaches have been proposed for MANET multicasting, but either suffer from low Packet Delivery Ratio (PDR) or high overhead and lack of scalability. In addition, reliable multicasting requires re-transmission of lost packets, which increases protocol overhead. Recently, Reinforcement Learning (RL) techniques have been successfully used in unicast routing to provide adaptive routing schemes. RL allows wireless nodes to make efficient routing decisions based on interaction with the environment while reducing the routing overhead compared to traditional routing approaches. In this paper, we investigate whether the same results hold when applying RL to multicasting in MANETs. We aim for two main performance criteria: ensuring 100% data delivery while at the same time reducing the total number of packets transmitted over the network. Based on these metrics, we evaluate existing RL-based multicast routing techniques and suggest promising approaches for reliable and efficient multicast routing protocols.