Non-alcoholic fatty liver disease (NAFLD) is the hepatic manifestation of the metabolic syndrome. Recent evidence has suggested the protective effects of honey consumption against the metabolic syndrome, but the association between honey intake and NAFLD is still unclear. We investigated how the consumption frequency of honey was associated with NAFLD in the general population. This was a cross-sectional study of 21 979 adults aged 20–90 years. NAFLD was diagnosed based on the ultrasound-diagnosed fatty liver without significant alcohol intake and other liver diseases. Diet information, including consumption frequency of honey, was assessed by a validated 100-item FFQ. OR with 95 % CI were calculated by the binary logistic regression model, adjusting for confounding factors identified by the directed acyclic graph. Overall, 6513 adults (29·6 %) had NAFLD. Compared with participants consuming ≤1 time/week of honey, the multivariable OR of NAFLD were 0·86 (95 % CI 0·77, 0·97) for 2–6 times/week and 1·10 (95 % CI 0·95, 1·27) for ≥1 times/d (Pfor trend = 0·90). The results were generally similar in subgroups of BMI at a cut-point of 24·0 kg/m2 (Pfor interaction = 0·10). In this large-scale study, consuming honey 2–6 times/week was inversely associated with NAFLD, whereas consuming honey ≥1 times/d had no association with NAFLD. These results need replication in other large-scale prospective studies.