Campaigns spend over $137 million on AI ads as deepfakes pose risks to consumers, voters and others, security experts warn
Realistic fakes, growing concerns
Advances in artificial intelligence have made it much easier to create convincing fake images, videos and audio. Experts warn that bad actors can use them to target people with scams and spread disinformation.
Deepfakes imitate the appearance or voice of private individuals and public figures. They may include a caller sounding like a relative or colleague, or a politician appearing to ask voters for support or donations. Doppel co-founder Kevin Tian said that in one test, 92% of people thought an AI agent calling them was human; only 8% recognized the agent as AI.
Political ads and disinformation
A manipulated social media video in Texas purported to show a former student accusing Senate candidate James Talarico, a Democrat, of sexual misconduct, according to PolitiFact, a service of The Poynter Institute. By mid-October, political campaigns had spent more than $137 million on at least 220 AI-generated ads, some featuring candidate deepfakes, according to the Wesleyan Media Project, which tracks political advertising.
Project co-director Erika Franklin Fowler said spending on AI-generated ads was less than $50,000 in the 2024 cycle. She also said creators usually do not disclose when their ads were made with AI.
Fraud targets businesses, consumers
Raul Liive, a deepfake expert at identity verification company Veriff, said the speed and ease of creating and distributing deepfakes pose major risks. Veriff documented a 300% increase in deepfake-related fraud attempts against businesses this year compared with 2025.
Some scams use claims of medical emergencies, such as a caller asking for urgent help paying a medical bill. Veriff's research also found that consumers can have trouble recognizing deepfakes: in one test, 70% mistook a deepfake video of a female character for a real person.
Visual clues are fading
Older signs of fake content, such as a person appearing with six fingers or three hands, are becoming less common. Veriff found people still look for visual irregularities, including unnatural skin or unusual hair and teeth, but AI is increasingly removing those cues.
Kaiser said people may be unsure whether what they see is legitimate or whether donation requests are real. That uncertainty, Kaiser said, can undermine trust across the system.
Key points
- A test found that 92% of people thought an AI caller was human.
- Political campaigns spent more than $137 million on at least 220 AI-generated ads by mid-October.
- Veriff documented a 300% rise in deepfake-related fraud attempts against businesses compared with 2025.
- In a Veriff test, 70% of participants mistook a deepfake video for real.