Ad hoc networks represent a key factor in the evolution of wireless communications. These networks typically consist of equal nodes that communicate without central control, interacting in a peer-to-peer way. In such a network, an efficient and scalable data retrieval constitutes a challenging problem. Unstructured P2P networks avoid the limitations of centralized systems and the drawbacks of structured approaches, because they impose few constraints on topology and data placement, and support highly versatile search mechanisms. However their search algorithms are usually based on simple flooding schemes, showing severe inefficiencies. In this paper, in order to address this major limitation, we evaluate the adoption of a local adaptive routing protocol, suitable for a self-organizing ad hoc environment. The routing algorithm uses a simple reinforcement learning scheme (driven by query interactions among peers), in order to dynamically adapting the topology to peer interests. In our simulation this approach is able to dynamically group peer nodes in clusters containing peers with shared interests and organized into a small world network.