Tag clouds have become a frequently used interaction technique in web-based systems. Recently, different clustered presentation approaches have been suggested to improve usability and utility of tag clouds. In this paper we describe a modified layout strategy for clustered tag clouds and report the findings of an empirical evaluation of automatically clustered tag clouds with 22 participants for both specific and general search tasks. The evaluation showed that automatically clustered presentation performs as well as alphabetic layouts in specific search tasks and that clustered presentation is an improvement over random layout for general search tasks. Clustered tag cloud presentation also was preferred by a majority of users for general search tasks. High quality of the clustering was mentioned as key variable for usefulness of the approach in the qualitative interviews with the users.