With an increasing number of supported devices, usability evaluation of websites becomes a laborious task. Therefore, usability evaluation should be automated as far as possible. In this paper, we present a summative method for automated usability evaluation of websites. The approach records user actions and transforms them into task trees. The task trees are then checked for usability smells to identify potential usability issues. The approach was applied in two case studies and shows promising results in the identification of four types of usability smells.