Background and aims: Tolerance to the effects of alcohol is an important element in the diagnosis of alcohol use disorders (AUD); however, there is ongoing debate about its utility in the diagnosis AUD in adolescents and young adults. This study aimed to refine the assessment of tolerance in young adults by testing different definitions of tolerance and their associations with longitudinal AUD outcomes. Design: Prospective cohort study. Settings: Australia. Participants: A contemporary cohort of emerging adults across Australia (n = 565, mean age = 18.9, range = 18–21 at baseline). Measurements: Clinician-administered Structured Clinical Interview for DSM-IV Research Version (SCID-IV-RV) assessed for AUD criteria across five interviews, at 6-month intervals over 2.5 years. Tolerance definitions were operationalized using survey-type response (yes/no), clinician judgement (SCID-IV-RV), different initial drinking quantity and percentage increase thresholds and average heavy consumption metrics. AUD persistence was operationalized by the number of times AUD was present across the 2.5-year study period (n = 491), and new-onset AUD was operationalized as any new incidence of AUD during the follow-up period (n = 461). Findings: The (i) SCID-IV-RV clinician judgement [odds ratio (OR) = 2.50, P = 0.005], (ii) an initial drinking quantity threshold of four to five drinks and 50% minimum increase (OR = 2.48, P = 0.007) and (iii) 50% increase only (OR = 2.40, P = 0.005) were the tolerance definitions more strongly associated with any new onset of AUD throughout the four follow-up time-points than other definitions. However, these definitions were not associated with persistent AUD (Ps > 0.05). Average heavy consumption definitions of tolerance were most strongly associated with persistent AUD (OR = 6.66, P = 0.001; OR = 4.65, P = 0.004) but not associated with new-onset AUD (Ps > 0.05). Conclusions: Initial drink and percentage change thresholds appear to improve the efficacy of change-based tolerance as an indicator for new-onset alcohol use disorder diagnosis in self-report surveys of young adults. When predicting persistent alcohol use disorder, average heavy consumption-based indicators appear to be a better way to measure tolerance than self-reported change-based definitions. © 2022 The Authors. Addiction published by John Wiley & Sons Ltd on behalf of Society for the Study of Addiction.