Seeing is No Longer Believing: A systematic Review of Authenticity, Trust, and Behavioural Responses to AI-Generated Visual Content

Abstract

Generative AI is making visual content increasingly realistic, which makes it harder to assess its source, credibility, and legitimacy. Existing research offers limited insight into how the perceived authenticity of AI-generated visual content is formed and how it influences trust and behaviour. This study systematically reviewed 26 peer-reviewed research papers published between 2024 and 2026, integrating them through descriptive statistics, narrative synthesis, and thematic analysis. Based on the review, this study conceptualises the relationship between claims, evidence, and risk as ‘consequential authenticity’, which is used to examine whether visual evidence can adequately support a specific claim, and whether the extent of such support is commensurate with the potential risks. This perspective distinguishes between source recognition, perceived authenticity, trust and behavioural responses, whilst linking perceptual judgements to situational legitimacy. The study aims to provide a systematic synthesis of a fragmented body of research, offer a conceptual framework to guide future empirical work, and identify directions for future research, including cross-cultural, longitudinal and behavioural designs that can further clarify the situational boundaries of authenticity judgements.