Brazil’s 2026 presidential election is becoming an important experiment in how artificial intelligence can reshape democratic politics. With the October election approaching, AI-generated political influencers, which are digital personalities who look and speak like real people, are attracting millions of views on TikTok, Instagram, and other platforms.

One prominent example is ‘Dona Maria,’ an elderly Black Brazilian woman who frequently criticizes President da Silva and discusses subjects ranging from food prices to foreign policy. She appears to be an ordinary citizen expressing political frustration, but she does not exist. She was created using generative AI by Daniel Cristiano dos Santos, a ride-share driver from Rio de Janeiro. Some of her videos have received millions of views, and researchers estimate that her content may have reached tens of millions of people.

The rise of AI influencers in Brazil presents novel legal and political challenges. From a legal standpoint, laws prohibiting the unauthorized duplication or representation of real people don’t apply because the AI generated influencers are not duplicates of any one person. AI influencers cannot be convinced, persuaded, or even educated the way humans can, which raises interesting and potentially troubling questions about their limitations or lack thereof.

Their rise demonstrates how generative AI can blur the distinction between ordinary political speech, organized campaigning, and manufactured public opinion. Dona Maria is part of a broader phenomenon. Researchers with Brazil’s Observatório IA nas Eleições identified at least 18 artificial political influencers currently in use. They take the form of voters, commentators, financial experts, and supposed community figures.

Bloomberg reporting notes that these synthetic personalities can be duplicated, redesigned, and used to produce virtually unlimited variations of political messages without sleeping, demanding salaries, or maintaining a consistent real-world identity. Eleven of the 18 identified avatars lacked clear disclosure that they were AI-generated.

The problem is especially serious because much of the material is misleading. Research cited by CNN Brazil found that 78 percent of content distributed by the identified artificial personalities contained false or distorted claims about politicians, government programs, or democratic institutions.

Brazil nevertheless has some of the world’s strongest election rules governing artificial intelligence. The Superior Electoral Court, or TSE, requires synthetic campaign material to carry clear AI labels. It also prohibits certain new AI-generated depictions of candidates and public figures during the crucial period beginning 72 hours before voting and ending 24 hours afterward. Platforms can also be required to remove prohibited synthetic material.

Enforcement is far more difficult than writing the rules. An AI avatar cannot itself be fined or prosecuted. Authorities must identify whoever created, financed, or distributed it, even when accounts are anonymous or characters have been copied across numerous platforms.

Generative AI makes persuasive political content extremely inexpensive, allowing individuals, campaigns or potentially foreign actors to operate networks of apparently independent ‘citizens.’ Instead of one famous deepfake changing an election, future manipulation may consist of thousands of artificial personalities gradually shaping what voters believe other people think.

Brazil’s experience may provide a preview of elections elsewhere. If its relatively aggressive regulations struggle to control synthetic influencers, countries with weaker election laws could face even greater problems. The 2026 Brazilian election is showing that the next challenge for democracy may not merely be determining whether a video is fake it may be determining whether the person trying to influence voters ever existed at all.