Alcohol use disorders (AUD) are often comorbid with other disorders with high levels of impairment, which is of relevance for the development and the progression of the disease. Evidence shows that AUD varies greatly with regard to its aetiology, which might lead to distinct clinical representations with important implications for treatment. The current study aimed to apply latent class analysis (LCA) techniques to investigate how comorbidity patterns in AUD vary with regard to specific explanatory factors. A Swiss community sample of N=439 individuals with AUD was subjected to LCA in order to find empirical AUD subtypes of comorbid psychiatric conditions. The subtypes were further validated based on a range of external criteria, including clinical and psycho-social factors as well as treatment variables. A three-class solution of empirical subtypes of AUD comorbidity (low, depressive-anxious, and drug-dependent antisocial) provided the best fit to the data. The three AUD subtypes showed homogeneous comorbidity patterns but varied along dimensions of psycho-social risk factors, consumption patterns and consequences as well as treatment history. Our findings provide strong evidence that AUD in non-treated samples can be described as a multidimensional disorder in terms of its comorbidity structure with distinct etiological factors and important consequences for treatment.