The advent of indirect measures, such as the Implicit Association test (IAT), has stimulated interest in implicit cognitions that may automatically steer addictive behaviours such as alcohol abuse. Counter-intuitively, recent IAT research has demonstrated that alcohol is implicitly associated with negative valence, regardless of the level of alcohol consumption. However, because the IAT is susceptible to extrapersonal contamination, this study examined whether previous findings reflect contamination of IAT effects by negative extrapersonal knowledge rather than personal associations with alcohol. Implicit alcohol associations were measured with a personalized alcohol-IAT, designed to reduce extrapersonal contamination. Whether alcohol associations measured with the personalized IAT would predict drinking behaviour above the variance explained by self-reported alcohol-related expectancies and attitude was examined. In contrast to previous findings with the IAT, the personalized IAT yielded positive associations. Moreover, positive alcohol associations predicted drinking behaviour above self-reported alcohol expectancies and attitudes, demonstrating the incremental validity of the personalized IAT. The present findings support the hypothesis that previous findings with the alcohol-IAT at least partly reflect negative extrapersonal alcohol-related knowledge, and that implicit alcohol associations are positive rather than negative.