Background: This experimental study evaluated the interobserver reliability and accuracy of pre-operative digital templating for humeral head size, stem size and neck angle for total shoulder arthroplasty. Methods: Twenty-five patients underwent a total shoulder arthroplasty with a single prosthesis. Four independent, blinded surgeons (two experienced shoulder surgeons and two PGY-6 fellows) used pre-operative radiographs and templating software to generate templates of the humeral head, stem and neck for each patient. Interobserver reliability was calculated using weighted kappa (k) analysis. Accuracy was assessed by comparing templates to actual implant sizes. Results: Interobserver reliability was fair to substantial (k ¼ 0.26 to 0.71) for head size, fair to substantial (k ¼ 0.39 to 0.72) for stem size and slight to fair (k ¼ 0.16 to 0.34) for neck angle. Templated head size, stem size and neck angle had accuracies of 53%, 77% and 68% within one size variation, respectively. Experience did not affect accuracy (p ¼ 0.11 to 0.48). Conclusions: Digital templating is not a useful guide for pre-operative surgical planning and should not be used to select a prosthesis.