BRASÍLIA, Brazil – Meta has denied political bias in Brazil’s presidential race. The denial follows an order from the country’s Superior Electoral Court requiring equal treatment for President Luiz Inácio Lula da Silva and opposition candidate Flávio Bolsonaro.
The court issued its order on October 15. It covered how Meta distributes political posts, recommends accounts and displays search results. It also addressed the company’s use of artificial intelligence (AI) during the election. Meta faces a reported fine of 500,000 Brazilian reais per day if it fails to comply.
Lula’s campaign had accused Meta of restricting its content while allowing different treatment for Bolsonaro-related material. Meta blamed technical problems, anti-fraud checks and efforts to follow Brazilian election rules.
However, the court’s preliminary assessment did not establish that the AI-related failure was deliberate. Unequal results can raise concerns about election fairness, but they do not automatically prove intentional political interference.
What the electoral court ordered
On October 15, TSE Justice Marco Aurélio Bellizze ordered Meta to ensure equal treatment for both presidential candidates. Agência Brasil reported that the order covered post distribution, account recommendations and search results.
The case followed complaints from Lula’s campaign. It alleged that Meta restricted some content related to Lula while allowing certain Bolsonaro-related content to appear. The campaign also raised concerns about differences in the company’s AI image-generation tools.
The court had already sought answers from Meta. On October 10, Bellizze gave the company 48 hours to explain alleged differences in content distribution, recommendations and geographic restrictions.
The later order required Meta to apply the relevant rules equally. However, an order to ensure fairness is not the same as a finding of deliberate misconduct. The available reporting does not establish that the judge found Meta had intentionally manipulated the election in Bolsonaro’s favor.
Meta points to technical problems
In a statement published on October 16, Meta rejected claims that it was interfering in Brazil’s election. The company said its systems do not favor political ideologies and that its rules apply equally to candidates and parties.
Meta described several separate incidents to support its position.
The company said an anti-fraud review mistakenly disabled Lula’s campaign advertising account for just over an hour on September 23. The review involved an advertisement that appeared to resemble a government social-benefit program, according to Meta.
Meta said it corrected the restriction. It also said the incident did not block Lula’s main Facebook page or Instagram account.
Another problem affected video-view counts on October 7. Meta said the issue lasted about five hours and affected users in more than 100 countries. According to the company, the fault changed how view counts appeared but did not affect content distribution. Meta also said Bolsonaro’s campaign reported the problem.
A third incident occurred on October 9. A software update temporarily stopped advertisers from promoting existing Instagram feed posts. Meta said advertisers could still create new ads through other methods. Existing ads continued to run, and the company said it resolved the problem by reversing the update.
These explanations come from Meta. Independent technical records would help establish the incidents’ full effects and whether they affected the campaigns differently.
Internal messages raise more questions
On October 16, Folha de S.Paulo reported on internal Meta messages about restrictions affecting Lula’s campaign and recommendations involving Bolsonaro-related accounts. Reuters also reported on the newspaper’s findings.
According to Folha, Lula’s official Instagram account could not promote existing feed posts for almost 24 hours from October 9. The campaign estimated that about 1.3 million people missed its advertisements because of the restriction.
That figure is the campaign’s estimate. It has not been independently established as a measure of lost reach, and it does not show how the restriction affected voter behavior.
The newspaper also reported that internal messages described problems with recommendations and searches involving Bolsonaro-related accounts. Some cases reportedly involved Instagram searches and WhatsApp channel searches.
Meta told Folha that different problems caused the incidents. The company cited technical faults and an error that it said it had corrected. It denied giving either candidate preferential treatment.
Bolsonaro’s campaign, in comments reported by the newspaper, said it had not experienced problems that disrupted the candidate’s advertising account.
The reported messages provide more detail about the dispute. However, they do not by themselves prove that Meta deliberately tried to influence the election. The original messages, technical records and correction timeline would help establish what happened.
AI tools produced different results
The dispute over AI attracted particular attention. Brazil’s election rules restrict platforms from using AI to rank, recommend, suggest or prioritize candidates and campaigns.
Lula’s campaign said Meta’s AI tools rejected some requests to generate images of the president while producing images involving Bolsonaro in similar cases. Agência Brasil reported examples of these different responses and the campaign’s complaint to the electoral court.
Meta said it had temporarily stopped its AI tools from recommending candidates or generating images of them in Brazil to comply with election rules.
The company blamed a gap in its safeguards for the different results. Meta said its system blocked prompts that clearly named candidates but did not always recognize a candidate when a common first name appeared alongside an election number.
Meta said it introduced additional safeguards to address the problem. However, independent tests would be needed to establish whether those changes worked consistently across prompts and services.
AI image generation and political recommendations also involve different functions. One tool creates an image in response to a request. Another system may rank or promote political accounts. Both can raise election-law concerns, but investigators need different evidence to assess each one.
Equal treatment remains the central question
Brazil’s presidential runoff is scheduled for October 25. The case has raised questions about how digital platforms enforce election rules when their systems affect political advertising, search visibility and AI-generated content.
Meta’s denial does not prove that every system treated the candidates fairly. Equally, technical failures and inconsistent results do not automatically prove deliberate political bias.
Investigators must examine each incident separately. They need to establish what happened, how long it lasted, which accounts were affected and whether Meta applied its rules consistently.
The court’s order sets a clear requirement: Meta must treat the candidates equally under the relevant rules. Whether the company has met that requirement depends on its conduct after the order and the evidence available to the court.
The reported daily fine applies if Meta fails to comply. It does not, on its own, establish that the company has violated the order.
Several questions remain. Did the incidents affect the campaigns differently in practice? Do the reported internal messages accurately reflect how Meta’s systems worked? And what evidence will the court use to assess compliance before the runoff?
For now, the dispute centres on both the results produced by Meta’s systems and the company’s explanations for them. Brazil election platform oversight will remain important as the court assesses whether both candidates receive the equal treatment required by law.
Reporting Credit: The Superior Electoral Court’s public statement and Meta’s October 16, 2026, response were consulted for the court’s stated priorities and the company’s explanations. Agência Brasil’s reporting helped establish the reported scope of the order and the earlier request for answers. The internal-message allegations are attributed to Folha de S.Paulo, with Reuters coverage used as secondary corroboration. Direct responses from both campaigns, the complete court decision and independent technical records remain necessary for further verification.
