We evaluated the association between urinary arsenic and the seroprevalence of total hepatitis A antibodies (total anti-HAV: IgG and IgM) in 11 092 participants aged ⩾6 years using information collected in the US National Health and Nutrition Examination Survey (2003–2012). Multivariate logistic regression models evaluated associations between total anti-HAV and total urinary arsenic defined as the sum of arsenite, arsenate, monomethylarsonate and dimethylarsinate (TUA1). Effect modification by self-reported HAV immunization status was evaluated. Total anti-HAV seroprevalence was 35·1% [95% confidence interval (CI) 33·3–36·9]. Seropositive status was associated with higher arsenic levels and this association was modified by immunization status (P = 0·03). For participants that received ⩾2 vaccine doses or did not know if they had received any doses, a positive dose-response association was observed between increasing TUA1 and odds of total anti-HAV [odds ratio (OR) 1·42, 95% CI 1·11–1·81; and OR 1·75, 95% CI 1·22–2·52], respectively. A positive but not statistically significant association was observed in those who received <2 doses (OR 1·46, 95% CI 0·83–2·59) or no dose (OR 1·12, 95% CI 0·98–1·30). Our analysis indicates that prevalent arsenic exposure was associated with positive total anti-HAV seroprevalence. Further studies are needed to determine if arsenic increases the risk for incident hepatitis A infection or HAV seroconversion.