Background: Accumulated studies have shown that low expression of 25-hydroxyvitamin D [25(OH)D] was significantly associated with the risk of non-alcoholic fatty liver disease (NAFLD). However, the exact causality is still unknown. The aim of this study was to investigate whether levels of 25(OH)D are associated with risk of NAFLD, using a two-sample Mendelian randomization (MR). Methods: Data from a recent large vitamin D genome-wide association study (GWAS) on 417,580 Europeans were utilized, and the largest published histology-based NAFLD GWAS study (1,483 cases and 17,781 healthy controls) for genetic variants predicted to cause NAFLD were searched. All genetic datasets for the MR analyses were obtained using publicly available summary statistics based on individuals of European ancestry from the MR-Base and NHGRI-EBI GWAS Catalog database. Inverse-variance weighted (IVW) MR approach was used to estimate causal effects in the main analysis, complemented by 4 additional methods to control for pleiotropy. Sensitivity analyses were conducted to verify whether heterogeneity and pleiotropy can bias the MR results. Results: The MR analysis did not provide strong evidence for the causal association of circulating 25(OH)D with NAFLD by IVW method (OR = 0.746, 95%CI 0.517–1.078; P = 0.119). The results were consistent using four other MR methods. Sensitivity analysis using all different analytical approaches yielded similar results. There was no evidence for pleiotropy (MR-Egger intercept: −0.0003758, P = 0.970). The replication process also showed consistent results using IVW method ( P = 0.710). Conclusion: This study indicates that serum 25(OH)D levels did not possess an obvious effect on the risk of NAFLD. The associations in previous studies may be due to residual confounding or reverse causation.