This study investigates how gendered communication strategies in political self-representation influence voter perceptions, drawing on role incongruity theory and the double bind dilemma. Specifically, we explore (1) how self-representation strategies on social media — emphasizing communal or agentic traits — affect voter evaluations of competence, likability, and voting intentions, and (2) whether these effects differ by candidate gender and policy issue. A pre-registered online 2x2x2 between-subjects survey experiment, with a power-determined sample (N=3400), examined voter reactions to pre-tested AI-generated Twitter/X timelines of fictional candidates exhibiting communal or agentic traits through visuals. Hypotheses include that female politicians displaying agentic traits will be rated higher on competence but lower on likability, while male politicians exhibiting communal traits will score higher on likability but lower on competence. Furthermore, we hypothesize that counter-stereotypic strategies will have distinct impacts based on candidate gender, and these effects will vary between hard-policy (defense) and soft-policy (welfare) topics. This study addresses methodological gaps by emphasizing visual cues, which are critical for trait inferences yet often neglected in text-based analyses. Prior research suggests voters rely on visual information as primary indicators of gender, implying that previous text-focused findings may underestimate these effects (Coronel et al., 2021). This may clarify the current conflicting evidence regarding stereotype incongruity in candidate evaluations and its electoral consequences for women in politics (Rohrbach et al., 2022). In doing so, our findings contribute to broader discussions on gender and political leadership, examining whether the increased visibility of women in political leadership has shifted voter stereotypes.