The contents follow each other in the infinite scrolling of the networks. In a video, an avatar of an elderly woman criticizes the president () with an indignant speech filled with expletives. A photo simulates the senator () next to the banker and Careca do .
The presidential candidate (New) attacks the (Federal Supreme Court) with synthetically created characters representing the ministers. (Avante), also a pre-candidate for Planalto, presents simulacrums of angry men dressed in yellow and red, in a critique of polarization.
Five months from now, content produced with generative as a trench for political dispute inside and outside the official pre-campaign apparatus.
Survey by , a project created by Aláfia Lab and Data Privacy Brasil, mapped 137 contents that deal with political issues using AI on networks between December 2025 and February 2026.
Of the monitored content, only 27% signaled the use of AI with the use of watermarks, captions or texts in the image. The remaining content was published without any warning, including posts made by politicians. Most of the content circulated through , and X.
“We identified an increase in the volume of synthetic content and widespread use by politicians and parties. Furthermore, the content is much more realistic and aesthetically believable”, assesses Matheus Soares, content coordinator at Aláfia Lab.
In March, the (Superior Electoral Court) for the use of AI in campaigns. Among the measures adopted is the ban on electoral content produced by AI 72 hours before and 24 hours after each round of voting.
The court also maintained the determination that advertisements must indicate the existence of synthetic content and inform which technology was used. Another rule provides for the banning of false, apocryphal or automated profiles whenever there are practices that could compromise the process.
Despite this, the company faces a complex scenario when analyzing content on the border between satire, meme and disinformation.
In the first months of 2026, publications triggered legal disputes, with allegations of misinformation or deepfake (which simulates people’s voices and images and creates a false reality in photos and videos).
The PT filed two lawsuits with the TSE in February questioning synthetic videos posted by the PL and party parliamentarians, including Flávio Bolsonaro, which portrayed Lula as a devil, showed the president in a prison uniform and associated him with corruption at the INSS.
The party returned to court after publishing videos of an elderly woman created by artificial intelligence who criticizes the Lula government. Lawyers maintain that the page spreads misinformation.
The “Dona Maria” videos gained social media engagement, some of them with millions of views. Report revealed that the page was created by app driver Daniel Cristiano, resident of Magé (RJ). He claims to have no connection with political parties.
A series of AI-created videos published by former Minas Gerais governor Romeu Zema also gained traction. The series, called “Os Intocáveis”, features avatars of STF ministers such as Dias Toffoli and Alexandre de Moraes.
Gilmar reacted and sent Moraes a criminal report against Zema and asked that he be investigated in the fake news investigation. The procedure is confidential.
The pre-candidate’s videos inspired content with similar aesthetics in the states. In Pernambuco, councilor Thiago Medina (PL) published one that satirizes the former mayor (PSB), called “the heir” in an AI animation. The same in Mato Grosso do Sul, in posts by deputy João Henrique Catan (Novo) against governor Eduardo Riedel (PP).
In Bahia, where the electoral dispute is polarized between (PT) and (União Brasil), media ecosystems orbit around the two candidacies on the networks, with no relation to the official profiles of the pre-candidates and with extensive use of artificial intelligence.
In recent months, there have been favorable decisions for both sides of the dispute in the Electoral Court. Magistrates ordered the removal of content made with AI with offensive potential, decided to suspend pages and adopted measures to identify anonymous profile administrators.
In March, the Electoral Court accepted the PT’s representation and ordered the removal of the @jerolandiabahia profile on Instagram, which had a critical tone towards the governor.
In the same week, the state’s Regional Electoral Court accepted União Brasil’s action and ordered the anonymous profiles @prefakesalvador and @acmmasterbahia on Instagram to be suspended. At least one of the videos used deepfakes produced with AI with content critical of the former mayor of Salvador.
In both cases, the judges pointed to laws that prohibit anonymity in electoral propaganda broadcast over the internet and also the resolution that prohibits the use of manufactured or manipulated content to disseminate untrue or out-of-contextual facts.
In Alagoas, the TRE ordered the suspension of a support profile for the senator’s candidacy () after identifying signs of irregular propaganda using deepfake to attack JHC (PSDB). The video simulates false statements attributed to the former mayor of Maceió.
Coordinator of digital platforms and markets at Data Privacy Brasil, Carla Rodrigues assesses that the intensity of synthetic content production on networks is expected to increase until the elections. And he highlights that one of the main challenges will be accountability for potentially offensive content.
“The big challenge is to understand this balance between the speed of production and dissemination of synthetic content. And also to verify whether preliminary decisions will be faster, especially when there is informational damage in a short period of two months, which is the electoral period”, he states.
