When Fake Feels Real: Typologies and Biases in AI Visual Content

This paper explores how AI-generated images (AIGIs) deceive users in social media environments by triggering cognitive biases and emotional reactions. Based on the Grounded Theory (GT) methodology, the analysis resulted in a typology of AIGIs, and user comments ($\mathrm{n}=9.082$) collected from Facebook pages operated within the Meta ecosystem using systematic and theoretical sampling. Findings reveal that visual stimuli such as nostalgic, spiritual, and empathetic images generate high engagement and foster misplaced trust. Instead of providing a full methodological account, this extended abstract highlights the most relevant insights from the GT analysis, specifically, how distinct image types correlate with emotional deception, and draws attention to the urgent need for research, regulation, and public awareness concerning the spread of AI-generated visual misinformation.