When Seeing Is No Longer Believing: Reddit Users’ Discernment of AI-Generated Visual Content

As generative artificial intelligence (AI) increasingly blurs the boundary between real and synthetic visual media, determining whether online content is authentic has become more difficult. This study examines how Reddit users collectively make sense of suspected AI-generated visual content, including images and videos. Rather than assessing detection accuracy, we examine how users judge whether visual content is real or AI-generated. Analyzing 89,697 Reddit comments from AI-focused communities, we identify four major forms of authenticity reasoning: real-world plausibility, visual artifact inspection, source-based assessment, and intuition. Real-world plausibility and visual artifact inspection were the most frequently used strategies, while source-based assessment appeared less often. Authenticity judgments also did not necessarily converge over time. Instead, discussions often diverged as users introduced competing interpretations. These findings suggest that evaluating visual AI-generated content is not simply an individual detection task, but a platformed sensemaking process shaped by users’ reasoning, intuition, and interactions with others.