Implied Authenticity Effect? The Impact of Explicit Labels on AI-Generated Content

This study investigates how labeling AI-generated content (AIGC) influences users’ perceptions of authenticity in social media environments. Motivated by emerging global regulations mandating the disclosure of AI-generated media, we designed a pre-registered survey experiment to test two main questions: (1) Do AI labels reduce the perceived authenticity of AI-generated images? (2) Does exposure to labeled content affect perceived authenticity in unlabeled images (a potential spillover effect)? The pre-analysis plan and materials are available on OSF (see link below). We conducted a survey experiment with a German sample (N = 877) in which participants were randomly assigned to one of three groups: a control group without labels, a process-based label group (“AI-generated”), or a harm-based label group (“Misleading”). Participants viewed twelve Instagram-style posts, six of which contained AI-generated or AI-altered content. Perceived authenticity was measured by asking whether the depicted events actually occurred. Our results show that both labeling strategies significantly reduced perceived authenticity of AI-generated images, with average reductions of about 0.27 standard deviations. We also find evidence of an implied authenticity effect: exposure to labeled content slightly increased perceived authenticity in unlabeled images (about one-fifth the size of the direct labeling effect). Exploratory analyses indicate that internet skills and age are associated with perceived authenticity differences. By embedding labels in Instagram-style posts, our study increases ecological validity compared to earlier work focused on headlines or generic stimuli. Situated in Germany, a frontrunner in digital platform regulation, the findings provide rare non-U.S. evidence on AI labeling and contribute directly to ongoing policy debates.