Self-report measures directly ask respondents to report their mental content, such as thoughts and feelings. By contrast, implicit measures aim to assess thoughts and feelings using performance indicators (for example, response times, error rates and response frequencies) under conditions that favour automatic processing. Implicit measures are now widely used in psychological science and beyond, because they are assumed to be superior to self-reports in various ways. In this Perspective, we argue that, despite the enthusiasm for implicit measures, self-reports are most often the better measurement option. First, the use of implicit measures is often based on mistaken assumptions about the disadvantages of self-reports. Second, self-reports have advantageous characteristics that are currently unmatched in implicit measures. We call for a more sophisticated use of self-reports and for caution when using implicit measures in basic and applied research.