The Impact of AI Attitudes on Detecting AI-Generated Political Content

The rise of generative artificial intelligence (AI) raises critical questions about users’ ability to distinguish AI-generated content from human-generated content, particularly in the political domain. This study investigates the relationship between subjective attitudes towards AI and the accuracy of detecting AI-generated political content. In an experiment involving political science students (n = 60), participants were asked to identify definitions of “political values” generated by the ChatGPT model and to assess their degree of ideological bias. The results revealed that, on average, only about half of the participants correctly identified the artificial origin of the definitions. However, the group of “optimists”, who hold a positive view of AI development, demonstrated significantly higher detection accuracy compared to the “neutral” group. Paradoxically, it was the “neutral” respondents who were significantly more likely to rate AI-generated content as completely objective and devoid of ideological bias. This suggests that a neutral, rather than a skeptical, stance towards AI may create a blind spot for potential ideological influence, as such content does not undergo sufficient critical reflection. The study contributes to understanding how user attitudes mediate the perception of political content in the age of generative AI.