Story perspectives

AI Deepfake of Pelosi Cuts Voter Support, Fact-Check Ineffective

5/6/2026

Summary

Pelosi Deepfake

The study found AI-generated deepfake videos of Rep. Nancy Pelosi reduced voter support.
Over 3,000 U.S. and Dutch adults viewed either a genuine address or a deepfake—Pelosi delivering an extremist Capitol-riot statement or Sybrand Buma giving an anti-immigrant monologue.
The realistic deepfake increased processing fluency, causing participants to accept false content despite suspicion.
Fact-checking did not repair reputational damage, prompting authors to suggest testing deepfakes in live campaigns.

Summary

Politicians' Image Fades

By week’s end, participants’ negative feelings toward the politicians had largely faded, showing a short-term emotional impact that receded after exposure.
Voters in the United States and the Netherlands reacted in almost identical ways, indicating that susceptibility to AI-generated political fakes crosses cultural and systemic differences.
Pre-view media-literacy warnings produced almost no measurable impact.
Post-view fact-checks lowered perceived realism yet failed to repair the politicians’ reputational damage.

History

Historical Context

In early 2016 the Russian-funded Internet Research Agency began a systematic social-media campaign that created thousands of fake American personas, bought Facebook ads and posted pro-Trump, anti-Clinton content, spending about $100,000 on more than 3,500 ads. The 2020 DOJ indictment of the IRA set a precedent for modern digital deception, now seen in AI-generated deepfakes of politicians.

1 of 3

Pelosi Deepfake

The study found AI-generated deepfake videos of Rep. Nancy Pelosi reduced voter support.
Over 3,000 U.S. and Dutch adults viewed either a genuine address or a deepfake—Pelosi delivering an extremist Capitol-riot statement or Sybrand Buma giving an anti-immigrant monologue.
The realistic deepfake increased processing fluency, causing participants to accept false content despite suspicion.
Fact-checking did not repair reputational damage, prompting authors to suggest testing deepfakes in live campaigns.

1 / 3