One of the prominent, by now seminal, paradigms in the research tradition of cognitive dissonance (Festinger, 1957) is the free-choice paradigm developed by Brehm (1956) to measure choice-induced preference change. Some 50 years after Brehm introduced the paradigm, Chen and Risen (2010) published an influential critique arguing that what the paradigm measures is not necessarily a choice-induced preference change, but possibly an artifact of the choice revealing existing preferences. They showed that once the artifact is experimentally controlled for, there is either no or very little evidence for choice-induced preference change. Given the prominence of the paradigm, this critique meant that much of what we thought we knew about the psychological process of cognitive dissonance might not be true. Following the critique, research using the paradigm applied various corrections to overcome the artifact. The present research examined whether choice truly changes preferences, or rather merely reflects them. We conducted a meta-analysis on 43 studies (N = 2,191), all using an artifact-free free-choice paradigm. Using different meta-analytical methods, and conceptually different analyses, including a Bayesian one, we found an overall effect size of Cohen’s d = 0.40, 95% confidence interval (CI) [0.32, 0.49]. Furthermore, we found no evidence for publication bias as an alternative explanation for the choice-induced preference change effect. These results support the existence of true preference change created by choice.