In Brazil, artificial intelligence is playing a growing role in political campaigns, particularly in the upcoming 2026 presidential elections. Deepfakes, which are manipulated videos or audios, are being used by supporters of candidates like Luiz Inácio Lula da Silva and Flávio Bolsonaro. These deepfakes, often lacking disclosure of AI use, have accumulated millions of interactions on social media platforms, even though electoral rules prohibit such manipulations.
Beyond deepfakes of actual politicians, AI is also creating entirely fictional influencers and voters. On TikTok, for instance, numerous videos feature AI-generated individuals expressing support for or opposition to candidates, with some presenting themselves as beneficiaries of social policies or anti-corruption advocates. A study found at least 50 such videos simulating interviews and testimonials, garnering 7.4 million views, and 18% of them failed to disclose their AI origin.
The Tribunal Superior Eleitoral (TSE) in Brazil has issued resolutions to regulate AI use in elections. These rules forbid deepfakes and prohibit AI systems from ranking, recommending, or showing preference for candidates. However, enforcement is challenging, especially when content is produced by third parties with no direct ties to campaigns. The TSE has engaged with major tech companies like X, OpenAI, and Meta to address instances where AI models have recommended candidates, urging them to review and correct vulnerabilities in their systems.
Experts highlight that these AI-generated characters, like the fictional "Dona Maria" who criticizes the government, pose a significant threat to democratic debate. They can be customized with specific appearances, accents, and political stances to appeal to different segments of the electorate, potentially amplifying polarization. The ability to simulate real people makes these AI-generated influencers highly persuasive, as they can foster a sense of representation and significantly impact voter behavior, according to Rosana Hermann of Conversar I.A. da USP.
Despite the advanced use of AI in campaigns, a recent Latam Pulse survey indicated that 77.5% of Brazilians would not be swayed by a public endorsement from a digital influencer. While 44.6% believe influencers will have "some influence" and 26.8% anticipate "much influence" in the election, 72.3% reported never changing their political opinion due to an influencer. This suggests a potential disconnect between the perceived impact of AI-driven influence campaigns and the actual effect on voters.
