The Alzheimer disease 8 (AD8) is a brief informant-based measure that distinguishes individuals with very mild dementia and mild cognitive impairment from those with normal cognition. The aim of this study was to establish the validity, reliability, and discriminative properties of the Korean version of the AD8 (K-AD8). Evaluation was made on 155 patient-informant dyads. The K-AD8 scores with the Clinical Dementia Rating (CDR) and performance on neuropsychologic tests were compared. Internal consistency of the K-AD8 was significant (Cronbach alpha=0.88). The K-AD8 had strong correlation with CDR (Spearman rho=0.76). Concurrent validity was strong with the K-AD8 scores correlating with CDR domains and performance on neuropsychologic tests. The K-AD8 was found to have excellent test-retest reliability (weighted kappa=0.81) and good interrater reliability (intraclass correlation coefficient=0.89). The area under the receiver operator characteristic curve was 0.88, suggesting good discrimination between nondemented individuals and those with cognitive impairment. To distinguish CDR 0.5 from CDR 0, in particular, the K-AD8 (cutoff score of 2) had a sensitivity of 68% and a specificity of 90%. In conclusion, the K-AD8 is a sensitive screening tool in detecting very early dementia, indicating that the AD8 could work very well in a variety of cultural entities.