Abstract The current paper introduces a novel feature of implicit association tests (IATs) by demonstrating their potential to change implicit attitudes. We assume that such changes are driven by associative learning mechanisms caused by carrying out an IAT task. Currently, evaluative conditioning appears to be the only widespread paradigm for changing implicit attitudes. An IAT task could provide an alternative. In two experiments, participants initially reacted to only one IAT task. Implicit preferences subsequently assessed with different implicit measures depended on the initial IAT task. This was shown for implicit self-esteem and for implicit attitudes towards well-known candy brands. Findings are discussed in relation to task-order effects in IATs.