Deepfake Videos Undermine Politician Reputations Even When Recognized as Fake

A four-wave experiment titled “Radical Right-Wing Political Deepfakes Can Successfully Delegitimize Targeted Political Actors” examined how AI-generated deepfakes affect voter attitudes in United States and Netherlands. AI tools now enable face swaps and voice cloning, creating “deepfakes” that can portray politicians saying extremist statements—a threat to democratic elections. Over 3,000 adults viewed either a genuine address or a manipulated clip in which U.S. Representative Nancy Pelosi appeared to sympathize with Capitol rioters, or Dutch Christian Democratic leader Sybrand Buma delivered an extremist anti-immigrant speech.

Psypost

Participants completed three surveys over a week in 2021. After a baseline questionnaire, they were randomly assigned to (1) view the deepfake, (2) view it after a media-literacy warning, or (3) view it followed by a fact-check refuting the false claims. The survey measured changes in support for the targeted politicians and belief in video authenticity.

Psypost

Even when participants correctly identified the clips as fake, support for Pelosi and Buma fell, especially among prior supporters. No shift occurred in political beliefs, and the reputational dip vanished after a week. Fact-checking lowered perceived realism but did not restore reputations; media-literacy warnings had negligible impact. Repeated exposure did not create an “illusion of truth.”

Psypost

The deepfakes used extreme rhetoric and visual glitches, making deception easier to spot. Voice actors supplied the audio, allowing attentive viewers to notice unnatural speech. Researchers note that subtler alterations may bypass detection, and that testing during live campaigns would better reflect real-world dynamics.

Psypost

Michael Hameleers stressed that “the vulnerability to artificial media transcends cultural borders,” citing similar reactions in both the polarized U.S. system and the consensus-driven Dutch context. He called for “future tests during live political campaigns” to gauge deepfake effects amid competing news coverage.

Psypost

The study did not examine subtle deepfakes nor track long-term attitude changes beyond the experimental week, leaving the durability of reputational harm in natural settings uncertain.

Psypost

Researchers recommend real-time experiments during upcoming elections to monitor voter reactions to authentic digital propaganda. Findings could shape fact-checking protocols and media-literacy interventions as AI-generated misinformation evolves.

Psypost

• “Artificial intelligence can be used to generate deceptive videos that damage a politician’s reputation, even when viewers suspect the footage is fake.” — Michael Hameleers, University of Amsterdam

Psypost

• “Even though people correctly suspected the videos were fake, their opinions of the politicians still dropped.” — Michael Hameleers, University of Amsterdam

Psypost

• “Yet, those exact same fact-checks completely failed to reverse the emotional damage done to the politicians’ reputations.” — Michael Hameleers, University of Amsterdam

Psypost

• “The deepfakes successfully damaged the reputations of both Pelosi and Buma.” — Michael Hameleers, University of Amsterdam

Psypost