Seeing Is Not Believing: Study Reveals AI-Generated Videos Erode Trust in Authentic Content

Summary

Researchers at MIT Media Lab published findings from a two-phase study (N=100) examining how exposure to realistic AI-generated videos affects people's perception of authentic videos. The research found that even when synthetic content is clearly disclosed, participants exposed to AI-generated videos first reported significantly increased doubt about subsequently viewed authentic videos, reduced judgment confidence, greater perceptual disruption, and lower social connectedness compared to a control group that viewed only human-generated content.

The study challenges the effectiveness of disclosure and detection as sole safeguards against the broader impacts of generative video technology. The researchers argue that beyond addressing deception risk, the field must develop design strategies that prioritize 'perceptual safety'—addressing the experiential and psychological consequences that realistic AI videos impose on media consumers, regardless of transparency labels.

Editorial Opinion

This research exposes a critical blind spot in how the tech industry approaches synthetic media: disclosure-based approaches treat the problem as one of deception when the real issue is corrosion of trust itself. The psychological residue of watching convincing AI videos appears to linger, poisoning perception of all subsequent media regardless of authenticity. As generative video becomes indistinguishable from reality, the industry must shift from 'can we detect this?' to 'how do we design systems that preserve human confidence in media ecosystems fundamentally altered by synthetic content?'