Recent meta-analytical and multi-country studies employing vignettes, conjoint experiments, and explicit preference measures conclude that citizens, on average, prefer women political candidates (e.g., Schwarz & Coppock, 2022; Gothreau & Laustsen, 2024). And yet, women remain underrepresented in politics across the Globe. To better understand this obvious discrepancy, this project employs a well-validated experimental paradigm from the psychological sciences—the reverse correlation (RC) technique (Dotsch & Todorov, 2012; Schmitz et al. 2024)—to further scrutinize if citizens really prefer women candidates. The RC technique is a two-stage method that evokes participants’ visual representations of social categories, here political candidates. In Stage 1 (N=1,000), American participants are exposed to an array of visually distorted versions of the same androgynous face and asked to choose which face they prefer as politician. After repeating this choice many times, participants’ chosen faces are averaged to create a composite face that serves as a visual representation of their preferred politician. In stage 2, another sample of American participants (N=500), naive to Stage 1’s face-selection procedure, rate how well the composite faces resemble “a woman” or “a man” illuminating if participants from Stage 1 hold gendered mental representations of “political candidates”. Results provide an original and unobtrusive estimate of citizens’ preferences for candidate gender that is arguably less influenced by response and social desirability biases than existing work. Further, the RC technique allows for descriptively comparing whether explicitly stated preferences for women candidates are (mis)aligned with the way preferred politicians are represented “in the mind’s eye”.