AI in Political Campaigns

How Campaigns Use AI Today

Political campaigns have adopted AI tools across every function. At the most basic level, campaigns use LLMs to draft fundraising emails, social media posts, and press releases — tasks that previously required teams of copywriters. The Republican National Committee ran an entirely AI-generated attack ad against President Biden in April 2023, using synthetic images of fabricated future scenarios to depict the consequences of his reelection.

More sophisticated uses involve voter targeting. AI models can analyze voter files, social media activity, and consumer data to predict which voters are persuadable and what messages will move them. This is an evolution of the microtargeting that Cambridge Analytica made infamous, but with far more powerful models. A campaign can generate thousands of personalized email or text variations, each tailored to the recipient's predicted concerns.

The most concerning use is in synthetic media: AI-generated images, audio, and video designed to mislead voters about a candidate's positions, actions, or character.

The Transparency Gap

The central problem is that AI-generated campaign content is often indistinguishable from human-created content, and most jurisdictions have no disclosure requirements. A voter who receives a personalized text message does not know whether it was written by a human volunteer, a campaign staffer, or an AI model that analyzed their social media profile.

Some states and countries have begun to act. In the United States, over 40 states introduced AI-in-elections bills between 2023 and 2024, with varying approaches. Some require disclosure labels on AI-generated political ads. Others ban deepfakes of candidates within a certain number of days before an election. The EU's AI Act classifies political AI applications as 'high risk,' requiring transparency and human oversight.

Enforcement remains the fundamental challenge. A deepfake video posted anonymously on social media and shared through private messaging groups is extremely difficult to trace to its creator, even if creating it violates the law.

The Liar's Dividend

There is a paradox at the heart of AI and elections. As AI makes it easier to create fake content, it also makes it easier for real content to be dismissed as fake. This is what researchers call the 'liar's dividend' — the ability of politicians caught on tape saying or doing something damaging to claim, with increasing plausibility, that the evidence is an AI fabrication.

This double-edged effect means that AI is not just a misinformation tool — it is a trust erosion tool. Even when no AI is involved, the mere possibility of AI generation gives dishonest actors a ready-made excuse. Solving this requires not just detection technology but robust provenance systems that can prove an image, audio file, or video is authentic.