In this paper, we describe a method of adaptive holistic 3D visualization of arbitrarily large datasets. The method, called AVE, permits to visualize entire datasets by the use of the most appropriate interface selected automatically from a number of available interfaces. The choice of an interface is based on the dataset properties. Different types of interfaces enable presenting the dataset categorized according to multiple criteria. The method is most suitable for abstract multidimensional datasets. Examples provided in this paper relate to visualization of data generated by indexing search engines.