The study finds widespread susceptibility to deepfakes, with nearly 60% of viewers believing deepfakes are real and real videos only about half the time identified as authentic, cutting across political affiliation and demographics.

Deepfake videos reached equal or higher credibility than real footage, and even confident viewers often could not reliably distinguish them from authentic videos.

Experts argue policy should go beyond individual responsibility to platform accountability, including provenance labeling, watermarking, digital identity measures, faster disinformation detection and takedowns, and legal frameworks for deliberate synthetic-media manipulation in elections.

Participants misclassified deepfakes as real 58.7% of the time, and only 16.4% correctly labeled them as fake, indicating progress in realism but poor detection compared with two years earlier.

There is a call for platform-level safeguards alongside public vigilance in consuming political information.

A UVU-Herbert Institute study shows AI-generated deepfake media can influence public opinion as powerfully as real media, potentially affecting elections at all levels.

The study highlights concrete examples, such as AI-generated deepfake ads targeting a Texas Hill race and an AI-generated reading of old tweets used by political operatives.

Researchers warn deepfakes can be indistinguishable from real content, underscoring the need for citizens to verify information before voting.

By April 2026, 31 states had enacted laws addressing deepfakes in political communications, typically restricting near elections and requiring disclosures.

UVU presented the findings on June 18, 2026, at an event organized by the Emerging Tech Policy Lab and the Herbert Institute.

Detection cannot rely on visuals or audio alone; skepticism, verification, and media-literacy are essential defenses against disinformation.

In experiments with 632 participants, a fictional ballot initiative and four videos showed that voting intentions shifted after viewing, with similar influence from both real and synthetic videos.