The resilience of social media users against Generative AI-based visual manipulation: A survey among Hungarian Facebook users

Abstract

Generative AI technologies, such as deepfakes and other synthetic media, introduce significant challenges to cybersecurity and digital information integrity. These technologies enable highly realistic and contextually convincing visual content, making detection increasingly challenging for both end-users and automated systems. This study explores three main objectives: (1) assessing users' ability to recognize AI-generated images, (2) analyzing the impact of demographic factors—particularly age—on detection accuracy, and (3) examining the relationship between recognition performance and trust in social media platforms. A survey of 318 Facebook users revealed an average recognition accuracy of 63.57%, which is significantly below the expected benchmark of 70%. Younger users (18–25) demonstrated higher accuracy (69.08%) than older users (26+), while higher platform trust correlated with lower detection performance (56.47% vs. 66.32%). A qualitative analysis further identified four dominant recognition strategies: anomaly-based detection, intuitive-affective evaluation, over-perfection cues, and experiential reasoning. These findings underscore the pressing need for enhanced media literacy, critical thinking skills, and adaptive cybersecurity measures to combat AI-driven misinformation and protect digital ecosystems.

Ähnliche Arbeiten