The aim of this study was to evaluate comparatively five methods for automating mesh generation (AMG) when used to mesh a human femur. The five AMG methods considered were:​ mapped mesh, which provides hexahedral elements through a direct mapping of the element onto the geometry;​ tetra mesh, which generates tetrahedral elements from a solid model of the object geometry;​ voxel mesh which builds cubic 8-node elements directly from CT images;​ and hexa mesh that automatically generated hexahedral elements from a surface definition of the femur geometry. The various methods were tested against two reference models:​ a simplified geometric model and a proximal femur model. The first model was useful to assess the inherent accuracy of the meshes created by the AMG methods, since an analytical solution was available for the elastic problem of the simplified geometric model. The femur model was used to test the AMG methods in a more realistic condition. The femoral geometry was derived from a reference model (the “standardized femur”) and the finite element analyses predictions were compared to experimental measurements. All methods were evaluated in terms of human and computer effort needed to carry out the complete analysis, and in terms of accuracy. The comparison demonstrated that each tested method deserves attention and may be the best for specific situations. The mapped AMG method requires a significant human effort but is very accurate and it allows a tight control of the mesh structure. The tetra AMG method requires a solid model of the object to be analysed but is widely available and accurate. The hexa AMG method requires a significant computer effort but can also be used on polygonal models and is very accurate. The voxel AMG method requires a huge number of elements to reach an accuracy comparable to that of the other methods, but it does not require any pre-processing of the CT dataset to extract the geometry and in some cases may be the only viable solution.