Objective To evaluate the diagnostic accuracy of observers with different levels of experience in reading F-18-sodium fluoride (NaF) PET/CT images for the diagnosis of bone metastases in prostate cancer (PCa) patients. Methods Nine observers with varying NaF PET/CT experience, ranging from no experience to 2000+ examinations, evaluated 211 NaF PET/CT scans from PCa patients participating in one of four prospective trials. Each observer evaluated each NaF PET/CT on a patient level using a trichotomous scale: M0 (no bone metastases), Me (equivocal for bone metastases) and M1 (bone metastases). Subsequently, a dichotomous evaluation was conducted (M0/M1). The final diagnosis was retrieved from the original study. For each observer, ROC curves and the diagnostic accuracy were calculated based on dichotomous and trichotomous scales; in the latter case, Me was first regarded as M1 and then M0. Results Across all experience levels, the sensitivity, specificity and accuracy using the dichotomous scale ranged from 0.81 to 0.89, 0.93 to 1.00 and 0.91 to 0.94, respectively. Employing the trichotomous scale, novice and experienced observers chose Me in up to 20 vs. 10% of cases, respectively. Considering Me as M0, the sensitivity, specificity and accuracy ranged from 0.78 to 0.89, 0.95 to 1.00 and 0.91 to 0.95, respectively. Considering Me as M1, the sensitivity, specificity and accuracy ranged from 0.86 to 0.92, 0.71 to 0.96 and 0.77 to 0.94, respectively. Conclusion Novice observers used the equivocal option more frequently than observers with NaF PET/CT experience. However, on the dichotomous scale, all observers exhibited high and satisfactory accuracy for the detection of bone metastases, making NaF PET/CT an effective imaging modality even in unexperienced hands.