Background Many observational studies focus on the relationship between Nonalcoholic fatty liver disease (NAFLD) and bone mineral density (BMD). However, the conclusions are controversial and the causal relationship between NAFLD and BMD remains unclear. Method A bi-directional two-sample Mendelian randomization (MR) analysis was performed to investigate the potential causal links between NAFLD and BMDs. We applied genetic variants as instrumental variables obtained from the Genetic Factors for Osteoporosis (GEFOS) dataset and several published genome-wide association studies (GWASs). We obtained summary statistics for heel (H) BMD (n = 426,824), femoral neck (FN) BMD (n = 32,735), lumbar spine (LS) BMD (n = 28,498), ultra-distal forearm (UF) BMD (n = 21,907), and total body (TB) BMD (n = 56,284) from some GWAS meta-analyses. Additionally, the NAFLD GWAS included 377,988 individuals of European ancestry which consist of 4,761 NAFLD cases and 373,227 control cases. We used inverse variance weighted (IVW), four supplemental methods, and several sensitivity analyses to estimated and cross-validate the potential causal relationship in the present MR analysis. Results The sensitivity analyses do not find any violation of the MR assumptions. We found that NAFLD has no causal association with H-BMD (beta − 0.017; 95%CI -0.0458,0.0117; p = 0.2461), FN-BMD (beta − 0.0166; 95%CI -0.1592,0.1259; p = 0.8191), LS-BMD (beta − 0.021; 95%CI -0.1475,0.1055; p = 0.7446), UF-BMD (beta − 0.0524; 95%CI -0.1726,0.0679; p = 0.3935), TB-BMD (beta − 0.0596, 95%CI -0.1236,0.0044; p = 0.0678). Similarly, reverse MR analysis provided little support for a causal effect of BMDs on NAFLD. Conclusion This MR study found no evidence to support a bi-directional causality between NAFLD and BMD.