Extant research on technology acceptance has devoted considerable attention to the relationship between behavioral intention (BI) and system use (SU) over time. However, the empirical results have been mixed—studies have found BI to both influence and not influence SU. Studies that examined the BI→SU relationship have employed different research models and research designs. Prior research models have examined direct effects on SU, indirect effects on SU through BI, the moderator effects on BI→SU, and the mediating role of BI in BI→SU. Studies have employed different types of respondents, information technologies, geographic regions, voluntariness, and measurement designs. This study proposes that such differences in research models and designs contribute to the mixed results for BI→SU, and reports a meta-analysis of findings reported in 113 prior studies along with a critical review of the BI→SU relationship. While no significant differences were found in the results for the BI→SU relationship across various research design characteristics, this study finds that the measurement of SU has been conflated with future and current or past behavior, direct and indirect effects on SU impact BI→SU, BI is modeled to fully and partially mediate the effects of other variables on SU, and moderators for the BI→SU relationship may be necessary. This study identifies several directions for future research and underlines the need to rethink the link between BI and SU.