Although self-rated health (SRH), a subjective measure of overall health status, associates with metabolic abnormalities, studies on the relationship between SRH and non-alcoholic fatty liver disease (NAFLD), a hepatic manifestation of metabolic syndrome, are limited. In this study, we evaluated whether or not SRH predicts the risk of incident NAFLD. This cohort study was performed in a sample of 148,313 Korean adults free of ultrasound-diagnosed NAFLD at baseline with annual or biennial follow-up for a median of 3.7 years. SRH and NAFLD were measured at baseline and follow-up visits. NAFLD was determined based on the ultrasound-diagnosed fatty liver without excessive alcohol consumption or any other cause. Hazard ratios with 95% confidence intervals were estimated via a parametric proportional hazards model. During 522,696.1 person-years of follow-up, 23,855 individuals with new-onset NAFLD were identified (incidence rate, 45.6 per 1,000 person-years). After adjustments for possible confounders including total calorie intake, sleep duration, and depressive symptoms, the multivariate-adjusted hazard ratios (95% confidence intervals) for incident NAFLD comparing good, fair, and poor or very poor SRH to very good SRH were 1.06 (0.97-1.14), 1.18 (1.09-1.27), and 1.24 (1.13-1.37), respectively. This association of SRH with incident NAFLD remained significant after accounting for changes in SRH and confounders during follow-up and was similar across clinically relevant subgroups. In a large-scale cohort study of apparently healthy Korean adults, poor SRH was independently and positively associated with incident NAFLD risk, indicating a predictive role of SRH as a health measure in NAFLD.