VeroRank study examines AI responses to 2026 Brazilian elections

A study by VeroRank analyzed how generative AI systems, including ChatGPT, Gemini, and Google AI Overview, respond to queries regarding the 2026 Brazilian presidential elections. Conducted between September 8 and 23, the research identified discrepancies in official data from the Superior Electoral Court (TSE) and a low rate of source citations.

While ChatGPT accounted for 65% of source consultations, it provided citations for only 20% of those instances, crediting only 14 out of every 100 sources read. In contrast, Gemini and AI Overview cited every source they consulted. The TSE website emerged as a leader in both consultations and citations.

Overall, journalism accounted for 53% of all citations, followed by official bodies at 17%, and social media or video platforms at 13%. The study highlights a gap between an AI reading a domain and actually providing it as a reference in a response.