Description
Generative artificial intelligence is increasingly used to evaluate political information, summarize candidate profiles, and generate strategic recommendations. As these systems become embedded within democratic information environments, they may begin to function as informal political gatekeepers that shape perceptions of political legitimacy and electoral viability. Yet little research has examined how AI systems evaluate political candidates or whether they reproduce existing inequalities in democratic representation.
This paper introduces the concept of algorithmic electability to explain how generative AI systems may reproduce assumptions regarding political leadership, legitimacy, and candidate viability. Drawing on an experimental study conducted across six major generative AI platforms, the analysis evaluates candidate profiles that vary across LGBTQ identity cues, gender presentation, and leadership style while holding qualifications constant. The findings indicate that AI systems frequently distinguish between competence and electability. Candidates who depart from conventional leadership expectations are often evaluated as equally qualified but less politically viable. The magnitude of these effects varies substantially across AI models.
The paper argues that generative AI should be understood not merely as an information technology but as an emerging political institution capable of influencing democratic representation. As AI systems become increasingly integrated into political communication and decision-making processes, understanding their role in shaping perceptions of political legitimacy becomes an important challenge for democratic governance.