The aim of this study was to determine the accuracy of registration and the precision of the resection volume in navigated hip arthroscopy for cam-type femoroacetabular impingement, using imageless and image-based registration. A virtual cam lesion was defined in 12 paired cadaver hips and randomly assigned to either imageless or image-based (three-dimensional (3D) fluoroscopy) navigated arthroscopic head–neck osteochondroplasty. The accuracy of patient–image registration for both protocols was evaluated and post-operative imaging was performed to evaluate the accuracy of the surgical resection. We found that the estimated accuracy of imageless registration in the arthroscopic setting was poor, with a mean error of 5.6 mm (standard deviation (. sd. ) 4.08; 95% confidence interval (CI) 4.14 to 7.19). Because of the significant mismatch between the actual position of the probe during surgery and the position of that probe as displayed on the navigation platform screen, navigated femoral osteochondroplasty was physically impossible. The estimated accuracy of image-based registration by means of 3D fluoroscopy had a mean error of 0.8 mm (. sd. 0.51; 95% CI 0.56 to 0.94). In terms of the volume of bony resection, a mean of 17% (. sd. 11; -6% to 28%) more bone was resected than with the virtual plan (p = 0.02). The resection was a mean of 1 mm deeper (. sd. 0.7; -0.3 to 1.6) larger than on the original virtual plan (p = 0.02). In conclusion, given the limited femoral surface that can be reached and digitised during arthroscopy of the hip, imageless registration is inaccurate and does not allow for reliable surgical navigation. However, image-based registration does acceptably allow for guided femoral osteochondroplasty in the arthroscopic management of femoroacetabular impingement