Background and purpose - In hip arthroplasty, acetabular inclination and anteversion-and also femoral stem torsion-are generally assessed by eye intraoperatively. We assessed whether visual estimation of cup and stem position is reliable. Patients and methods - In the course of a subgroup analysis of a prospective clinical trial, 65 patients underwent cementless hip arthroplasty using a minimally invasive anterolateral approach in lateral decubitus position. Altogether, 4 experienced surgeons assessed cup position intraoperatively according to the operative definition by Murray in the anterior pelvic plane and stem torsion in relation to the femoral condylar plane. Inclination, anteversion, and stem torsion were measured blind postoperatively on 3D-CT and compared to intraoperative results. Results - The mean difference between the 3D-CT results and intraoperative estimations by eye was -4.9° (-18 to 8.7) for inclination, 9.7° (-16 to 41) for anteversion, and -7.3° (-34 to 15) for stem torsion. We found an overestimation of > 5° for cup inclination in 32 hips, an overestimation of > 5° for stem torsion in 40 hips, and an underestimation < 5° for cup anteversion in 42 hips. The level of professional experience and patient characteristics had no clinically relevant effect on the accuracy of estimation by eye. Altogether, 46 stems were located outside the native norm of 10-20° as defined by Tönnis, measured on 3D-CT. Interpretation - Even an experienced surgeon's intraoperative estimation of cup and stem position by eye is not reliable compared to 3D-CT in minimally invasive THA. The use of mechanical insertion jigs, intraoperative fluoroscopy, or imageless navigation is recommended for correct implant insertion.