Seeing Is Not Believing: Realistic AI Videos Disrupt Confidence in Real Videos (www.media.mit.edu)

A recent study presented at the 2026 CHI Conference on Human Factors in Computing Systems reveals that emerging realistic AI-generated videos significantly undermine public confidence in authentic video content. Researchers from various institutions, including notable figures like Wolf and Maes, explored how viewers’ trust in video evidence diminishes as AI technology advances, highlighting a growing challenge for media credibility and information verification. This research is particularly significant for the AI/ML community as it underscores the societal implications of deepfake technology and similar advancements. The findings suggest that as AI-generated visuals become more indistinguishable from reality, individuals may increasingly view all video content—regardless of its authenticity—with skepticism. The study calls for urgent discussions around ethical guidelines and technological solutions to combat misinformation, emphasizing the need for robust verification systems to maintain trust in visual media and protect democratic discourse.