The extent to which color preferences are universal or individualized is a long-standing question in science. Here, we investigate the sources of variance in color preference, leveraging a novel generative AI framework to visualize individualized color preference models. We first developed a generative AI model of color space capable of creating any color. We then assessed the sources of variance in people’s color preferences using images generated by this model, revealing a substantial degree of individual variability. This variability was replicated in an analysis of color preferences collected by other authors (Schloss et al., 2015). Finally, we employed a novel reverse correlation methodology using our generative AI model to construct individual color preferences and show that participants were most sensitive to images generated from their idiosyncratic model. These findings reveal that color preferences are largely driven by the individual evaluating specific colors, rather than any shared properties of the stimuli.