A tool designed to manipulate perception without accountability
In the digital corridors of Brazilian political life, the Workers' Party has introduced an AI-generated version of 'Dona Maria' — a beloved popular character — to carry leftist messaging directly into the social media feeds of ordinary citizens. The move reflects a deeper transformation underway in democratic politics globally, where synthetic personas now stand in for human voices, blurring the line between authentic civic discourse and engineered persuasion. As Brazil approaches its political season, this experiment forces a reckoning with an old question in new form: when a voice speaks to the people, must there be a person behind it?
- PT-affiliated pages deployed an AI-generated character modeled on a proven popular persona, designed to deliver political messaging in the warm, familiar tone of everyday Brazilian conversation.
- The synthetic Dona Maria can post continuously, adapt in real time, and scale across platforms — capabilities no human influencer can match, and with none of the disclosure obligations they carry.
- Brazilian outlet Veja exposed the practice, spotlighting a regulatory blind spot where AI-generated political characters operate without the transparency requirements applied to paid human promoters.
- Other parties are watching: if one side normalizes synthetic political personas, the incentive to follow is immediate, threatening to flood electoral discourse with unaccountable artificial voices.
- Regulators and voters now face the urgent task of determining what transparency standards, if any, can be enforced before synthetic characters become the dominant medium of political persuasion.
In the months before Brazil's political season, pages aligned with the Workers' Party began circulating a new social media presence: an AI-generated version of Dona Maria, a character who had already earned a loyal following through her commentary on everyday life and politics. The PT-linked operatives saw in her popularity a template — and commissioned an artificial intelligence system to produce a left-leaning variant, one capable of speaking to voters in the same informal, relatable cadence that had made the original resonate.
The logic was strategic and unsentimental. A synthetic character could post without rest, adapt its messaging in real time, and reach across multiple platforms simultaneously — all while sounding like a neighbor discussing grocery prices or family struggles rather than a party broadcasting its platform. It was political communication dressed in the texture of ordinary life.
This was not an isolated experiment but part of a broader shift in Brazilian digital campaigning, where synthetic media — AI-written content, algorithmically optimized messaging, deepfakes — has become an increasingly deliberate instrument of narrative control. The Dona Maria project exemplified the trend: organic in appearance, engineered in origin.
The ethical fault line was clear. Voters scrolling past the character's posts had no indication they were reading machine-generated content shaped by party operatives. Unlike a paid human influencer, the AI persona carried no disclosure obligation — it could present itself as an independent citizen voice while functioning as a political tool.
When Veja and other Brazilian outlets began reporting on the practice, they framed it as a symptom of a larger crisis: electoral regulations had not kept pace with the technology. The deeper concern was not one character or one party, but the precedent being set. If synthetic political personas could operate freely, the future of democratic discourse might belong increasingly to characters without accountability, without reputation, and without the irreducible humanity that political speech has always, at least nominally, required.
In the months leading up to Brazil's political season, pages aligned with the Workers' Party began circulating a new character across social media: an AI-generated version of Dona Maria, a figure who had already proven her appeal to Brazilian audiences. The original Dona Maria had built a following through her commentary on everyday life and politics, becoming a recognizable voice in the digital landscape. The PT-linked pages saw an opportunity. They commissioned an artificial intelligence system to create their own iteration—a left-leaning variant designed to reach voters through the same informal, relatable tone that had made the original successful.
The move was neither subtle nor accidental. Political operatives understood that synthetic characters could bypass traditional media gatekeeping and speak directly to voters in spaces where they already spent time scrolling. An AI-generated persona could post constantly, adapt messaging in real time, and scale across multiple platforms without the constraints of a human operator. The character would deliver leftist political arguments wrapped in the familiar cadence of someone discussing bills, family, and the cost of living—the texture of ordinary conversation.
This tactic sits within a broader shift in how Brazilian political parties now approach digital campaigning. Rather than relying solely on human influencers or traditional advertising, campaigns increasingly deploy synthetic media: deepfakes, AI-written content, algorithmically optimized messaging. The Workers' Party's Dona Maria project exemplified this trend. It was not a one-off experiment but part of a deliberate strategy to control narrative space and amplify party messaging through channels that felt organic and peer-to-peer rather than top-down.
The creation of a left-wing Dona Maria raised immediate questions about authenticity and disclosure. Voters encountering the character on their feeds had no way of knowing they were reading words generated by a machine trained on patterns in text and image data. The character's posts appeared to come from a person with opinions, experiences, and a political perspective—but there was no person behind them. There was only code, trained weights, and the intentions of the party operatives who had built and deployed the system.
Brazilian media outlets, particularly Veja, began reporting on the practice, treating it as a symptom of a larger problem: the weaponization of AI in electoral politics. The concern was not merely that synthetic characters existed, but that they operated in a gray zone where political advertising regulations had not yet caught up with the technology. A human influencer paid to promote a party's message must disclose that relationship. An AI character generated by a party had no such obligation. It could speak as though it were an independent voice, a citizen sharing thoughts, when in fact it was a tool designed to manipulate perception.
The broader implications extended beyond one character or one party. If the Workers' Party had deployed an AI Dona Maria, other parties would likely follow. The technology was available, the platforms were willing, and the electoral incentives were clear. What emerged was a vision of political discourse increasingly mediated by synthetic actors—characters without accountability, without the constraints of human reputation, capable of reaching millions with messages tailored to micro-segments of the electorate. The question facing Brazilian regulators and voters alike was whether this represented the future of campaigning, and if so, what safeguards might be necessary to preserve some notion of authentic political speech.
A Conversa do Hearth Outra perspectiva sobre a história
Why would a political party go to the trouble of creating an AI character when they could just hire a human influencer?
Speed and scale. A human gets tired, takes days off, has limits on how many posts they can write. An AI character can post dozens of times a day, adapt messaging in minutes, and run across ten different platforms simultaneously. It's also cheaper in the long run.
But doesn't it feel dishonest? The voters don't know they're reading something a machine wrote.
That's exactly the problem. A human influencer has a reputation to protect. If they say something false, it damages their credibility. An AI character has no reputation—it's a tool. The party can discard it and create a new one if it becomes controversial.
So the original Dona Maria—was she real?
Yes. She had built an actual following by speaking in a way that resonated with people. The PT saw that and thought: we can replicate that appeal with our own version, one that says what we want it to say.
What happens if voters find out?
That's the risk. Right now there's no requirement to disclose that a character is AI-generated. But as more people realize this is happening, there will be pressure for transparency laws. The question is whether those laws come before or after the damage is done.
Damage to what?
To the idea that you're talking to another person when you're scrolling through your feed. Once that trust erodes, it's hard to rebuild.