BACKGROUND Alcohol dependence is a complex psychological disorder whose phenomenology changes as the disorder progresses. Neuroscience has provided a variety of theories and evidence for the development, maintenance, and severity of addiction; however, clinically, it has been difficult to evaluate alcohol use disorder (AUD) severity. OBJECTIVE This study seeks to evaluate and validate a data-driven approach to capturing alcohol severity in a community sample. METHODS Participants were non-treatment seeking problem drinkers (n = 283). A structural equation modeling approach was used to (a) verify the latent factor structure of the indices of AUD severity; and (b) test the relationship between the AUD severity factor and measures of alcohol use, affective symptoms, and motivation to change drinking. RESULTS The model was found to fit well, with all chosen indices of AUD severity loading significantly and positively onto the severity factor. In addition, the paths from the alcohol use, motivation, and affective factors accounted for 68% of the variance in AUD severity. Greater AUD severity was associated with greater alcohol use, increased affective symptoms, and higher motivation to change. CONCLUSIONS Unlike the categorical diagnostic criteria, the AUD severity factor is comprised of multiple quantitative dimensions of impairment observed across the progression of the disorder. The AUD severity factor was validated by testing it in relation to other outcomes such as alcohol use, affective symptoms, and motivation for change. Clinically, this approach to AUD severity can be used to inform treatment planning and ultimately to improve outcomes.