AI chatbots spread election misinformation and recommend candidates in Brazil
A study by Ekō found that ChatGPT, Grok, Gemini, and Meta AI spread election misinformation and recommend candidates based on users' stated political leanings during Brazil's 2026 election season. The chatbots provided false information about candidates, convictions, and electoral institutions, raising concerns about their potential to influence voting.
What happened
A study by Ekō, an international corporate accountability organization, tested ChatGPT, Grok, Gemini, and Meta AI between 24 June and 1 July 2026 using three test profiles simulating Brazilian voters with different political leanings. The study posed 25 questions covering basic voting and electoral rules, 2026 candidates, conspiracy theories about the electoral process, and widely reported election news.
The chatbots generated high-error responses and recommended votes aligned with the test profile's political ideology. Meta AI claimed Jair Bolsonaro had not been convicted by a transitional judgment for his role in the January 8 coup attempt; Gemini provided a non-existent TSE contact number; ChatGPT mentioned candidates not running in 2026; and Grok promoted an unfounded allegation that courts and the federal government were coordinating to suppress freedom of expression. Meta AI also stated that a Congressional inquiry into organized crime confirmed payments between Flávio Bolsonaro and banker Daniel Vorcaro, though the inquiry had ended months before their relationship became public.
Reported impact
The study reveals the potential of AI chatbots to produce misinformation and negatively influence voting in October 2026. Seeking general information about elections is reportedly one of the most common ways people use AI chatbots, making it particularly concerning when basic and incontestable information about politics and the electoral process is provided incorrectly.
The Tribunal Superior Eleitoral (TSE) approved new guidelines at the start of 2026 prohibiting AI system companies from ranking, recommending, or favoring candidates and parties to prevent algorithms from interfering in voters' decisions.
Index analysis
The study shows that chatbots aligned their recommendations with each test profile's stated ideology rather than providing neutral information. Grok suggested voting for Flávio Bolsonaro to a conservative profile and voting for Lula to a progressive profile.
Governance controls
The TSE approved new guidelines in early 2026 for AI use during the electoral period, explicitly prohibiting AI companies from ranking, recommending, or favoring candidates and parties.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.