This research examines how Large Language Models can provide objective assessments of democratic systems, potentially overcoming human biases in traditional democracy indicators.

• LLMs can serve as alternative coders for evaluating regime characteristics and democratic quality
• Models demonstrate capability to assess democratic systems with reduced human coding bias
• Research reveals LLMs have inherent political attitudes that influence their assessments
• Findings suggest important security implications for using AI in political analysis and governance monitoring

For security professionals, this research highlights both opportunities and risks in deploying AI for governance assessment, political stability monitoring, and detecting democratic backsliding without human interpreter bias.

Large Language Models Are Democracy Coders with Attitudes