Familiarity Matters: Examining the Effects of AI-Generated Images on Science Information Credibility

Abstract

Generative AI tools help illustrate complex scientific concepts, though concerns about accuracy persist. This study examined how AI-generated Instagram posts about CRISPR, with and without disclaimers indicating AI involvement, affect perceived credibility. In an online survey experiment with 742 U.S. adults, participants were randomly assigned to view content that varied by message frame (natural, neutral, or technology-focused) and the presence or absence of an AI disclaimer. Overall, AI disclaimers did not significantly affect credibility across frames. However, participants more familiar with AI perceived AI-disclaimed posts as more credible, while those less familiar perceived them as less credible.