Search engines are powerful tools to find information on the Web. However, they commonly return a lot of irrelevant documents when the users’ queries are not specific enough. To refine the scope of their searches, refinement terms are sometimes recommended to the users by query expansion systems (QES). A recent widescale survey has shown that users seldom include these terms in their queries. In this article, we propose a new user-centric approach for evaluating QES. The purpose of our approach is to assess the suggestive power of the suggested terms. Several existing QES are compared with respect to the proposed criteria.