There is an imperative to predict hazardous drinking among college students. Implicit measures have been useful in predicting unique variance in drinking and alcohol-related problems. However, they have been developed to test different theories of drinking and have rarely been directly compared with one another. Thus, their comparative utility is unclear. The current study examined five alcohol-related variants of the Implicit Association Test (IAT) in a sample of 300 undergraduates and sought to establish their predictive validity. Results indicated that the Drinking Identity IAT, which measured associations of “drinker” with “me,” was the most consistent predictor of alcohol consumption, alcohol problems, and alcohol cravings. It also had the highest internal consistency and test–retest reliability scores. The results for the Alcohol Excitement and Alcohol Approach IATs were also promising, but their psychometric properties were less consistent. Although the two IATs were positively correlated with all of the drinking outcome variables, they did not consistently predict unique variance in those variables after controlling for explicit measures. They also had relatively lower internal consistencies and test–retest reliabilities. Ultimately, results suggested that implicit drinking identity may be a useful tool for predicting alcohol consumption, problems, and cravings and a potential target for prevention and intervention efforts.