Both AI-generated and human influencers can correct misinformation: Investigating the effectiveness of corrections for polarized and non-polarized issues

Abstract

Misinformation in digital media environments poses a threat to democracy as it can bias beliefs, behavior, and policy. While social media influencers are often accused of contributing to the spread of misinformation, little is known about their potential role in combating it. In two pre-registered online experiments (N = 1004), we therefore investigated whether human and AI-generated influencers can effectively correct misinformation. The results show that influencers can successfully correct misperceptions on a politically non-polarized topic (mental health). On a politically polarized topic (climate change), however, only the correction of the AI-generated influencer significantly reduced the misperceptions. Overall, the results demonstrate that influencers can support the fight against misinformation and suggest that AI-generated influencers may be particularly effective in addressing more polarized political topics.