<p>Establishing the construct validity of virtual reality (VR) sports platforms is essential to ensure their effective and appropriate use with athletes. This study aimed to test: (1) the broad and fine construct validity of a commercially available VR golf game, and (2) whether real-world (RW) golf ability and videogame experience predicted VR golf performance. Thirty-three golfers and 32 non-golfers completed a 12-hole test (preceded by a five-hole familiarisation period), with their To Par score used as the performance measure. Self-reported current handicap, weekly golf hours, all-time videogame playing experience, and prior VR exposure was also recorded. A Welch’s <i>t</i> test found that golfers performed significantly better than non-golfers on the VR golf game (golfers <i>M</i> = 12.5 ± 7.5; non-golfers <i>M</i> = 27.5 ± 15.9), consistent with broad construct validity. VR golf performance was not predicted by current handicap, thereby failing to support fine construct validity. Weekly golf hours and prior VR exposure also did not predict VR golf performance, though videogame playing experience did, but only for the non-golfers group. Reduced motor correspondence (due to using a standard VR controller) and gamified features may explain the inability of the VR golf game to make fine-grained distinctions between golfers of varying abilities, whilst the minimal perception–action coupling afforded by videogames may account for the limited transfer of videogame experience to VR golf performance amongst golfers. Practitioners should ensure that VR sports platforms demonstrate (at least) broad construct validity before implementing them into the practices of their athletes.</p>