Can AI deepfakes change an election?

Yes. AI deepfakes can change an election, especially in a close race, but the bigger risk is often disruption and distrust rather than a guaranteed decisive swing.

They can affect elections in a few ways:

• Last-minute deception: A convincing fake released right before voting can leave little time to debunk.
• Voter suppression or confusion: Fake audio or video can spread false instructions about polling places, dates, eligibility, or how to vote.
• Targeted smear campaigns: Deepfakes can be aimed at small groups or local communities, making them harder to detect and correct.
• The “liar’s dividend”: Real evidence can be dismissed as fake, which can reduce accountability and public trust.

In real-world election cycles, audio deepfakes have already been used in attempts to influence voters, including efforts tied to turnout suppression. At the same time, major platforms and search engines have policies to label, demote, or remove synthetic media that misleads voters, and some places have passed laws requiring disclosure or restricting certain political deepfakes near elections.

The impact depends on timing, reach, the closeness of the race, and how quickly the fake is exposed. In many cases, deepfakes are more likely to deepen distrust and confusion than to single-handedly flip large numbers of votes.