How Humans and AI Detect Deepfakes Differently: Complementary Strategies for Human-AI Teaming

Deepfakes are AI-generated synthetic media that manipulate individuals’ lip movements, facial expressions, and voices, creating serious risks through deception, misinformation, and identity misuse. Although AI-based tools are essential for Deepfake detection, understanding and strengthening human detection ability is equally important. We compare human and AI attention allocation patterns, showing that they rely on distinct cues and reasoning strategies. We also examine physiological differences across levels of AI literacy to identify factors associated with superior human performance. Contrary to the assumption that AI consistently outperforms humans in Deepfake detection tasks, individuals with greater prior AI exposure may rival or exceed AI-based detectors in certain contexts. Eye-tracking and EEG evidence further suggests that humans use holistic, intuitive, and flexible visual-cognitive processing to detect subtle inconsistencies and artifacts overlooked by AI. These findings clarify how humans and AI process deceptive visual cues differently and inform trustworthy, adaptive human-AI teaming for Deepfake detection.