Vehicular ad hoc networks have recently been proposed as an effective tool for improving both road safety and the comfort experienced while driving. Vehicles may propagate information about potentially dangerous events such as lane changes or sudden slowdowns to vehicles in their vicinity. Moreover they can inform vehicles approaching from farther areas about accidents and possible traffic jams. In both cases, data must be routed to specific areas, along paths determined by the underlying road traffic conditions. In this paper we propose a novel approach to address this routing problem. First, we define a message propagation function that encodes information about both target areas and preferred routes. Second, we show how this function can be exploited in several routing protocols; and finally, we evaluate the effectiveness of our approach by means of simulation. Results highlight the good performance of our routing approach in sparse as well as in dense networks.