We argue and show empirically that constructs and measures of positive leadership styles, such as authentic, ethical, and servant leadership, are not veridical representations of leadership behaviors. Instead, these styles conflate behaviors with subjective evaluations of leaders. Labelling behaviors as, for example, “ethical” means evaluating leadership behaviors on positively valenced terms rather than describing these behaviors. Across four experiments, we show that positive leadership styles are outcomes that depend on non-behavioral, evaluative factors, such as information about a leader’s previous success or value alignment between leaders and followers. More importantly, the measures of these leadership styles create causal illusions by spuriously predicting objective outcomes, even when leader behaviors and other leader-specific factors are kept constant. Furthermore, these measures have predictive properties similar to those of a purely evaluative measure of leadership. In conclusion, our studies cast serious doubts on previous research claiming that positive leadership styles cause positive outcomes. Moreover, positive leadership style research is not only wrong but also practically futile because its constructs and measures are amalgams that do not isolate concrete and learnable behaviors. We call for a radical reorientation of leadership style research and sketch out options for more solid future research.