Objective: This study aims to explore the mechanisms of personality-targeted intervention effects on problematic drinking, internalizing and externalizing symptoms. Method: As part of a cluster-randomized trial, 1,210 high-risk students (mean age 13.7 years) in 19 London high schools (42.6% White, 54% male) were identified using the Substance Use Risk Profile Scale. Intervention school participants were invited to participate in personality-matched interventions by trained school staff. MacKinnon’s products of coefficients method was used to compare 3 complementary mechanism hypotheses, namely, whether early changes in (a) alcohol use, (b) internalizing and externalizing symptoms, or (c) personality during the 6 months postintervention accounted for intervention effects over 2 years. Results: Early intervention effects on drinking behaviors during the 6 months postintervention partially accounted for longer term intervention effects on the onset of binge drinking (95% confidence interval [CI] [-.349, -.062]) and drinking problems (95% CI [-.206, -.016]) over 2 years. Intervention effects on anxiety symptoms and conduct problems were partially mediated by early reductions in depressive symptoms (95% CI [-.013, -.001]; 95% CI [-.047, -.001]), and intervention effects on internalizing symptoms were also partially mediated by reductions in anxiety sensitivity (95% CI [-.003, 0]). Conclusions: 2-year intervention effects on problematic drinking were largely accounted for by early changes in drinking behaviors, and were not mediated by changes in mental health symptoms or personality risk factors. Early improvements in mood and anxiety sensitivity partially mediated longer term reductions in mental health problems.