The usage of virtual reality (VR) has been growing in many fields of research and therapy thanks to its immersive and gamified nature. Detection of the users’ subjective experience is thus essential for the effective personalization of content. Eye-tracking (ET) data and specifically gaze, in two-dimensional tasks, has been linked to value-based choices and emotional states. Therefore, here we aimed to develop a method for passive identification of subjective preferences based on ET data collected during a VR experience. For this purpose, we developed a naturalistic dynamic VR task where participants searched and looked at complex objects of pets and their control shapes that appeared in pre-defined locations in random order. At the end of the task, participants ranked their preference, valence, and arousal of the items they saw during the task. ET data was recorded using a built-in binocular eye-tracker within the VR headset. We found that the gaze behavior features of the median distance of gaze from the center of objects and the median gaze scan speed showed a significant interaction with object type (pets/shapes), as well as a significant positive relation to preference and valence rankings of pets. Our results suggest that these gaze behavior features could be used as passive biomarkers for detecting individual preferences and pleasantness, and in the future may enable successful personalization of VR content in real-time for various applications such as optimization of psychiatric diagnosis and treatment sessions.