Context: The concept that assessment of a person's health status without subsequent intervention has beneficial effects in itself has stimulated much interest in underlying psychological mechanisms, methodological implications and its public health potential. There have, however, been few experimental studies of assessment effects. Aim: To test the hypothesis that assessment in itself produces a reduction in hazardous drinking. Design: and setting Two conditions (group A, leaflet only and group B, leaflet and assessment but no intervention) of a four-arm randomized controlled trial with enrolment in March–April 2003. Participants: A total of 975 students (17–29 years) attending a primary health-care clinic completed a web-based Alcohol Use Disorders Identification Test (AUDIT) questionnaire. Of 599 who scored ≥ 8576 consented to follow-up and were included in the full four-arm trial, of whom 293 (153 women) were assigned to groups A and B. Intervention: Group A received an information leaflet at baseline. Group B received the information leaflet and 10 minutes of web-based assessment 4 weeks later. Measurements: Drinking frequency, typical quantity, heavy episode frequency, personal problems and academic problems. Findings: Baseline mean AUDIT scores were 15.0 (SD = 5.4) and 14.9 (SD = 5.0) in groups A and B, respectively. Twelve months after baseline, relative to group A, group B reported lower overall consumption (geometric means ratio 0.82, 95% CI: 0.68–0.98), fewer heavy drinking episodes (0.66, 0.47–0.91), fewer problems (0.81, 0.67–0.99) and lower AUDIT scores (beta = −1.63, −0.62 to −2.65). Conclusions: Brief assessment appeared to reduce hazardous drinking. Controlled trials that rely on assessment may therefore underestimate treatment effects. Limitations include the possibility of measurement artefact due to social desirability bias.