How Could AI Be Used to Influence Elections?

Synthetic Media Depicting Candidates

One of the most widely discussed ways AI could influence elections is through synthetic audio, image, or video content — sometimes called deepfakes — that realistically depicts a candidate or public figure saying or doing something they never actually said or did. Because generative AI tools have made producing this kind of realistic synthetic content considerably easier and cheaper than earlier manipulation techniques, election security researchers have flagged this as a significant emerging concern. A convincing fake video or audio clip released close to an election, when there may be limited time to debunk it before voters see and share it, is a scenario that has drawn particular attention from election officials and researchers.

Instances of AI-generated or manipulated political content have already appeared during actual election periods in multiple countries, illustrating that this is a live concern rather than a purely hypothetical one.

Large-Scale Disinformation and Automated Amplification

Beyond depicting specific individuals, AI can be used to generate large volumes of political disinformation content — false claims about candidates, policies, or voting procedures — more quickly and cheaply than would be possible through manual content creation alone. This content can then potentially be spread through automated bot accounts, which AI can make appear more natural and less obviously automated than earlier generations of bots, creating a false impression of grassroots public sentiment, momentum, or opposition around a candidate or issue.

This combination of large-scale content generation and automated amplification is a key reason researchers describe AI as changing the potential scale of election-related disinformation compared to earlier eras.

Personalized Political Messaging at Scale

AI can also enable highly personalized political messaging, sometimes called microtargeting, where tailored messages are directed at specific individuals or narrow demographic segments based on data about their characteristics, interests, or online behavior. While targeted political advertising predates modern AI, generative AI can make it more efficient to produce large numbers of tailored message variations, potentially allowing campaigns or bad actors to craft messaging optimized for very specific audiences at a scale that would be labor-intensive to achieve manually.

Researchers and election officials generally view this as raising both accountability questions — since narrowly targeted messages may receive less public scrutiny than broadly broadcast ones — and questions about transparency in political advertising more broadly.

Bottom Line

AI could be used to influence elections through realistic synthetic media falsely depicting candidates, large-scale AI-generated disinformation amplified by automated accounts, and highly personalized political microtargeting at scale. These are not purely hypothetical concerns — instances of AI-related election content have already been documented — which is why election officials, platforms, and lawmakers continue to develop responses to this evolving risk.

Go deeper

• What Are the Risks of AI-Generated Content Flooding Social Feeds?