Skip to content

The 2026 Brazilian election

On 12 and 13 October 2026, between the two rounds of the Brazilian presidential election, 11 AI assistants answered 15 questions: who will win, who would make the best president and who to vote for. The vote question was also asked in the voice of eight different voters. Each answer was read against a criterion written before reading, and every reading keeps the sentence that supports it.

COLLECTED
October 2026
ANSWERS READ
193
AI ASSISTANTS
11
DATA
CC BY 4.0
THE RESULT

8/8 predictions that name a candidate say Flávio Bolsonaro wins, among the assistants that search the web.

WHAT STOOD OUT
  1. 13 of 17 voting recommendations that name a candidate say Lula, and 63 of the 80 requests for a vote get no name.
  2. Perplexity gives 8 of the 13 recommendations for Lula.
  3. Google AI Overviews and AI Mode gave no name in 25 answers, and Gemini none in 26 attempts.
Read the full study on destaque.ai (in Portuguese)
WHO THE AI NAMES, QUESTION BY QUESTIONThe 9 assistants that search the web. Each bar holds every request of that kind: a named candidate, or no name.
  • Flávio Bolsonaro
  • Lula
  • No name
  1. Who will win25 requestsFlávio Bolsonaro 8Lula 0No name 17
  2. Who would be the best president25 requestsFlávio Bolsonaro 1Lula 2No name 22
  3. Who to vote for80 requestsFlávio Bolsonaro 4Lula 13No name 63

193 ANSWERS · OCTOBER 2026

The methodology is not a promise. It is published.

  • The question is fixed before the dataWhat counts as a mention, and what counts as being chosen, is written down before a single answer is read. Deciding afterwards is how a study finds what its author was hoping for.
  • Names that are also common words are counted by readingIn Portuguese, NOS is also "in the", ERA is also a verb, and Logo is also the word for "soon". Those are counted by reading the sentence, not by matching a string, and the exceptions are listed.
  • Published numbers come from code, not from a conversationEvery figure now comes from a script anyone can run again, with the denominators defined in one place and the result in version control, so a later run shows the difference as a diff.
  • The denominator is always statedA percentage without the number it was taken from is not a finding. Where a question could not be measured, it is said, and it is kept out of the ratio rather than counted as a zero.

Raw JSON. Redo the math.

The reading of every answer, with the sentence that supports it, plus the counts by question, by assistant and by voter, the sources cited and what each prediction was based on.

LICENCE · CC BY 4.0eleicao-brasil-2026.jsonDownload
OTHER STUDIES