Abstract Vehicular ad hoc network (VANET) routing protocols resort to clustering in order to optimize broadcast traffic flooding. Clustering schemes usually rely on rules which apply to each vehicle in order to reach a targeted organization in a VANET. Most of the literature works which evaluate clustering for VANET focus on performance analysis. However, with autonomous vehicles coming to roadways, more rigorous relationships will be required between clustering rules and the resulting organization, so as to anticipate road safety in a better way. We propose a formal description of the properties which are expected in a VANET, while considering the rules of a given clustering scheme. Using Event-B, we first present a description of the VANET, the vehicles movement and the traffic generated by both routing and application messages. Then, based on an Event-B model of a basic routing protocol of the literature, we describe how the specific rules of a clustering scheme can be modeled along with the properties expected in the resulting organization. Finally, we propose a validation process of the model. This paper aims at showing how our proposals have been applied to the Chain-Branch-Leaf scheme, although they can be adapted to any rule-based clustering scheme for VANET.