Binge (heavy episodic) drinking is common, but there is little knowledge on how this drinking pattern could be detected. This study compares three structured questionnaires among binge drinking middle-aged men. All 45-year-old men in the city of Tampere, Finland, were asked to fill in the Alcohol Use Disorders Identification Test (AUDIT). Based on the interview on their drinking the men were divided into non-binging moderate drinkers (n = 352), binging moderate drinkers (n =130), non-binging heavy drinkers (n = 10) and binging heavy drinkers (n = 63). The complete AUDIT, AUDIT-C (first three AUDIT questions inquiring quantity-frequency) and AUDIT-3 (the third binging-frequency question of AUDIT) in detecting binge drinking were compared. The complete AUDIT was effective in detecting binge drinkers by a cut-off score of ≥8 or ≥7. The optimal cut-off score for AUDIT-C was ≥6 and that for AUDIT-3 ≥2. The area under the curve (AUC) among all risky drinkers (binging moderate and binging heavy and non-binging heavy drinkers) for AUDIT was 0.824 (95% CI 0.789-0.859), for AUDIT-C 0.829 (95% CI 0.795 -0.864) and for AUDIT-3 0.779 (0.739-0.818). The complete AUDIT and its short versions are applicable in populations where binging is the dominant drinking pattern, but the cut-off scores should be tailored to individual cultures.