There is preliminary evidence that approach avoid training can shift implicit alcohol associations and improve treatment outcomes. We sought to replicate and extend those findings in US undergraduate social drinkers (Study 1) and at-risk drinkers (Study 2). Three adaptations of the approach avoid task (AAT) were tested. The first adaptation – the approach avoid training – was a replication and targeted implicit alcohol approach associations. The remaining two adaptations – the general identity and personalized identity trainings – targeted implicit drinking identity associations, which are robust predictors of hazardous drinking in US undergraduates. Study 1 included 300 undergraduate social drinkers. They were randomly assigned to real or sham training conditions for one of the three training adaptations, and completed two training sessions, spaced one week apart. Study 2 included 288 undergraduates at risk for alcohol use disorders. The same training procedures were used, but the two training sessions occurred within a single week. Results were not as expected. Across both studies, the approach avoid training yielded no evidence of training effects on implicit alcohol associations or alcohol outcomes. The general identity training also yielded no evidence of training effects on implicit alcohol associations or alcohol outcomes with one exception; individuals who completed real training demonstrated no changes in drinking refusal self-efficacy whereas individuals who completed sham training had reductions in self-efficacy. Finally, across both studies, the personalized identity training yielded no evidence of training effects on implicit alcohol associations or alcohol outcomes. Despite having relatively large samples and using a well-validated training task, study results indicated all three training adaptations were ineffective at this dose in US undergraduates. These findings are important because training studies are costly and labor-intensive. Future research may benefit from focusing on more severe populations, pairing training with other interventions, increasing training dose, and increasing gamification of training tasks.