Automatic clustering of semantic models allows a multilevel abstraction of the same reality and reduces design complexity. This paper is concerned with the development and application of automatic clustering of Entity-Relationship (E-R) diagrams and object models. This paper does not consider physical clustering which aims at improving performance of databases but deals with conceptual clustering whose objective is to facilitate the understanding of the database schema. We provide an automatization of conceptual schema clustering leading to a unification of past approaches. Automatization is achieved through the definition of semantic distances between concepts and the use of a clustering algorithm. For entity-relationship clustering, we define three different distances (visual, hierarchical and cohesive) depending on the semantic richness. Object model clustering is based on structural, semantic and communication characteristics of objects. Our approach has been implemented and applied to a great number of examples, leading to interesting results. Applications of automatic clustering in several areas and their potential as a communication, documentation and design tools for E-R diagrams and object oriented models are described and discussed.