This systematic review provides the first synthesis of mediating and moderating factors in the bidirectional relationship between anxiety and alcohol use and related problems. Six electronic databases were searched for longitudinal studies that assessed a mediator and/or moderator in the anxiety → alcohol or alcohol → anxiety pathway. Risk of bias was assessed with two quality assessment tools developed to assess biases pertaining to mediation and moderation studies, respectively. Of the 14,776 records identified, 55 were eligible, from which effects from 315 unique models were extracted. Effects included 30 mediation analyses, 258 moderation analyses, and 27 other complex analyses (e.g. multiple mediation or three-way moderation). Identified mediating and moderating factors were categorised in line with the biopsychosocial model, with subsequent subtheme classification (e.g. sex, drinking motives). Further to the narrative synthesis, seventeen moderation subthemes provided sufficient data for meta-analysis. Results of the meta-analysis of moderators suggest age (Fisher's z: 0.065 95% CI: 0.017, 0.113), externalising factors (Fisher's z: -0.186, 95% CI: -0.222, -0.150), perceptions of peer alcohol use (Fisher's z: -0.076, 95% CI: -0.119, -0.033), positive family experiences (Fisher's z: -0.081, 95% CI: -0.098, -0.064), and experimental manipulation of anxiety (Fisher's z: 0.242. 95% CI: 0.103, 0.382) significantly moderated the relationship between anxiety and alcohol. Narrative synthesis of other moderating subthemes and all mediation subthemes yielded inconsistent evidence which did not demonstrate conclusive moderated or mediated effects. Across studies, methodological quality was suboptimal, with future directions for research discussed.