An increasing number of studies have evaluated the association between ultra-processed foods (UPF) consumption and metabolic disorders. However, the association between UPF intake and non-alcoholic fatty liver disease (NAFLD) remains unclear. In this study, we analysed data from 6545 participants who were recruited in National Health and Nutrition Examination Surveys 2011–2018. UPF were defined in light of the NOVA food classification system and divided into quartiles based on its proportion of total weight intake. Complex logistic regression models were used to assess the association between UPF and NAFLD. Mediation analyses were conducted to reveal underlying mediators. We found that NAFLD patients consumed more UPF than controls (925·92 ± 18·08 v. 812·70 ± 14·32 g/d, P < 0·001). Dietary intake of UPF (% weight) was negatively related to the Healthy Eating Index-2015 score (Spearman r = −0·32, P < 0·001). In the multivariable model, the highest quartile compared with the lowest, the OR (95 % CI) were 1·83 (1·33, 2·53) for NAFLD (OR per 10 % increment: 1·15; 95 % CI: 1·09, 1·22; P for trend < 0·001) and 1·52 (1·12, 2·07) for insulin resistance (OR per 10 % increment: 1·11; 95 % CI: 1·05, 1·18; P for trend = 0·002). Mediation analyses revealed that poor diet quality, high saturated fat and refined grain intake partly mediated the association between UPF and NAFLD. In conclusion, high UPF intake was associated with an increased risk of NAFLD in US adults. Further prospective studies are needed to verify these findings.