Research on attention bias modification (ABM) for social anxiety disorder (SAD) is inconclusive, with some studies finding clear positive effects and other studies finding no significant benefit relative to control training procedures. In this meta-analysis, we assessed the efficacy of ABM for SAD on symptoms, reactivity to speech challenge, attentional bias (AB) toward threat, and secondary symptoms at posttraining as well as SAD symptoms at 4-month follow-up. A systematic search in bibliographical databases uncovered 15 randomized studies involving 1043 individuals that compared ABM to a control training procedure. Data were extracted independently by two raters. The Q statistic was used to assess homogeneity across trials. All analyses were conducted on intent-to-treat data. ABM produced a small but significant reduction in SAD symptoms (g=0.27), reactivity to speech challenge (g=0.46), and AB (g=0.30). These effects were moderated by characteristics of the ABM procedure, the design of the study, and trait anxiety at baseline. However, effects on secondary symptoms (g=0.09) and SAD symptoms at 4-month follow-up (g=0.09) were not significant. Although there was no indication of significant publication bias, the quality of the studies was substandard and wedged the effect sizes. From a clinical point of view, these findings imply that ABM is not yet ready for wide-scale dissemination as a treatment for SAD in routine care. Theoretical implications for the integration of AB in the conceptualization of SAD are discussed.