["Addiction, EarlyView. ", "\nAbstract\n\nBackground and aims\nMemory loss from alcohol use (‘blackout’) is a prevalent and costly indicator of substance‐related harm. While alcohol‐induced blackouts can be assessed retrospectively via self‐report, no tool exists to identify blackouts objectively in real time, hindering progress in research as well as intervention efforts. This study tested the feasibility and diagnostic accuracy of real‐time memory tasks in identifying alcohol‐induced blackouts in situ.\n\n\nDesign\nProspective diagnostic study.\n\n\nSetting\nUnited States.\n\n\nParticipants\nYoung adults (18–30 years of age; n = 63) reporting recurrent memory loss as a result of alcohol use were recruited from across the United States between December 2022 and January 2024.\n\n\nMeasurements\nParticipants completed a baseline survey and orientation interview, followed by 30 days of ecological momentary assessments (EMA). EMA included recall and recognition tests for visual stimuli during drinking events (index tests) and subsequent self‐reports of blackout (reference standard). At the end of the protocol, participants also completed a Timeline Followback interview of blackout events during the assessment period. Primary outcomes, which were specified prior to data collection, were (a) feasibility of memory task completion during drinking events and (b) diagnostic accuracy of memory tasks in identifying blackout at the event level. Data were analyzed using Bayesian logistic multilevel models.\n\n\nFindings\nOf the 63 participants included [mean age = 23.2, standard deviation (SD) = 3.3; 78% female, 51% White], 38 (60%) self‐reported a blackout during the assessment protocol. On average, participants completed 85% of memory tests prompted (SD = 16.42). Within days, both greater‐than‐average number of drinks [odds ratio (OR) = 1.74; 95% confidence interval (CI) = 1.41–2.19] and failure of 1 + recall memory test (OR = 15.53; 95% CI = 5.96–36.27) were associated with increased odds of blackout. In model‐predicted probabilities, blackout probability was ~0.01 if a person consumed their average number of drinks, ~0.21 if they had five more drinks than average and ~0.34 if they failed 1 + recall test. Participants self‐reported blackout on 39% of the days that they failed 1 + recall memory test (positive predictive value; 95% CI = 30–49); however, they reported not having a blackout on 92% of days that they correctly recalled all memory tests (negative predictive value; 95% CI = 89–95).\n\n\nConclusions\nObjective measures of alcohol‐induced blackout can be implemented in real‐life contexts. Failing a visual memory test while intoxicated is not necessarily indicative of blackout; however, correct recall indicates that blackout is highly unlikely.\n\n"]