AI Deepfake Political Ads Rise Ahead of US Midterms: How to Identify Fake Videos

Artificial intelligence (AI) is becoming an increasingly visible tool in political campaigning in the United States, with candidates using AI-generated videos, fictional scenarios and synthetic voices to criticise opponents and influence public opinion.

As the US midterm elections approach on November 3, 2026, the growing use of AI-generated political advertisements has raised concerns about misinformation, public trust and the distinction between political satire and factual claims.

Some campaigns openly use fabricated scenes rather than attempting to pass them off as genuine footage. However, even when advertisements include disclaimers identifying AI-generated content, viewers may still form impressions about candidates based on the messages and imagery presented.

AI-Generated Political Ads Target Candidates

Bruce Blakeman, a Republican candidate for governor of New York, has used AI-generated advertisements depicting Democratic leaders in fictional situations. One video portrays New York Governor Kathy Hochul and New York City Mayor Zohran Mamdani gardening together while supposedly discussing rising living costs. Another depicts Hochul supporting higher tolls and public transport fares.

Blakeman has defended the advertisements, arguing that disclaimers identifying the content as AI-generated make their nature clear. He has also compared the videos with the tradition of political parody.

The use of AI in election advertising extends beyond New York. According to the Wesleyan Media Project, a non-partisan organisation that tracks political advertising, candidates and their supporters have spent more than $80 million on over 164 AI-based advertisements during the current election season. Republican candidates and aligned groups accounted for approximately 80% of these advertisements, the project reported.

Other campaigns have used AI-generated imagery to ridicule political opponents. In Ohio, Democratic Senate candidate Sherrod Brown was portrayed as an unwelcome guest who had overstayed his welcome at a party. In Texas, Democratic Senate candidate James Talarico was depicted wearing a blue dress resembling the costume associated with Maria in The Sound of Music. Michigan Republican Senate candidate Mike Rogers was presented as a superhero.

These examples demonstrate how AI can be used not only to fabricate realistic scenes but also to create exaggerated and fictional portrayals intended to shape voters’ perceptions.

Concerns Grow Over Political Misinformation

Scott Babwah Brennen, director of New York University’s Center on Technology Policy and co-author of a report titled Political Slopaganda, has highlighted how AI-generated political content can influence audiences even when viewers understand that the footage is fictional.

Such advertisements may communicate broader judgements about a candidate’s character, values or competence without making a directly verifiable claim about a specific event.

A separate controversy in San Francisco involved California state Senator Scott Wiener, who introduced an AI chatbot targeting his opponent, Connie Chan, in a congressional race to succeed Nancy Pelosi. The chatbot reportedly answered around 5,000 questions and mocked aspects of Chan’s positions on property development, accent and naturalised citizenship.

After facing accusations that the tool was bigoted, Wiener withdrew it, acknowledging that the initiative had failed to achieve its intended purpose.

The incidents illustrate how AI-powered political campaigns can generate controversy over tone, fairness and the treatment of candidates, even when the content is not presented as authentic footage.

Are AI Deepfake Political Ads Regulated in the US?

The regulation of AI-generated political advertisements varies across the United States. According to the report, 31 states have enacted laws requiring labels on certain AI-generated political advertisements.

However, there is no single comprehensive federal law specifically regulating all political advertisements that use AI deepfakes. Existing rules differ in their scope and requirements.

The regulatory challenge involves balancing election integrity and transparency with protections for political expression, satire and parody. As AI tools become more accessible, policymakers face questions about how to ensure voters can identify synthetic content without unnecessarily restricting legitimate political speech.

How to Identify AI-Generated Fake Videos

AI-generated videos are becoming increasingly convincing, making it difficult to identify manipulated content through appearance alone. Nevertheless, several checks can help viewers assess whether a video may have been altered.

1. Check facial movements and expressions

Look for expressions that appear inconsistent with the conversation, unnatural blinking or changes in facial details between frames. However, these signs are not conclusive because compression and poor video quality can also produce visual distortions.

2. Examine hands and fingers

AI-generated footage can sometimes contain distorted fingers, unusual joint movements or hands that interact incorrectly with nearby objects. Examine the movement across multiple frames rather than relying on a single screenshot.

3. Inspect the background

Objects, signs and patterns may change shape, blur or appear inconsistent as the camera moves. Text on clothing, screens and signs may also shift between frames.

4. Listen for audio inconsistencies

Check whether the speaker’s lip movements match the audio. Unnatural pauses, unusual speech patterns or background sounds that do not fit the setting may warrant further investigation.

5. Check lighting and reflections

Shadows should broadly match their light sources, while reflections in mirrors and windows should remain consistent with the visible scene. Unexpected changes may indicate manipulation, although lighting errors can also occur in genuine recordings.

6. Verify the original source

This is one of the most important steps. Look for the original video, check the account that published it and see whether reliable news organisations or independent sources have verified the claims. Reverse image searches and video verification tools may also help locate earlier versions of the footage.

No single visual or audio clue can reliably prove that a video is fake. AI-generated content can appear realistic, while genuine footage may contain imperfections caused by editing, compression or recording conditions.

As AI becomes more common in political campaigning, checking a video’s source, context and supporting evidence is more dependable than judging its authenticity by appearance alone.