This study explored an optimal intensity measure for probabilistic seismic demand models of multiple components in cable-stayed bridges subjected to pulse-like ground motions. To achieve this goal, this study selected a cable-stayed bridge as a case study. Due to the limited number of recorded motions, artificial pulse-like motions were generated using an existing simulation method to obtain a reliable estimate of probability seismic demand models. For the subject bridge, seismic demand analyses were undertaken under a combined set of recorded (a total of 121 pairs) and artificial (a total of 121 pairs) ground motions. In total, seven structure-independent intensity measures and six types of engineering demand parameters representing maximum component responses were considered in this study. Peak ground acceleration (PGA) tended to be appropriate for the short-period components (i.e., pier and upper side of pylons), whereas peak ground velocity was found to be the optimal intensity measure for long-period components (i.e., lower side of pylons, cables, and displacement-related seismic demands). The lower side of pylons is more vulnerable than their upper side, and on the basis of efficiency, practicality, proficiency, sufficiency, and hazard computability, peak ground velocity appears to be the best intensity measure of the cable-stayed bridge subjected to pulse-like ground motions. However, PGA can be chosen as the ideal intensity measure for the piers.