President Luiz Inácio Lula da Silva has 40% of voter intentions in a first-round simulation for Brazil's 2026 presidential election, against 37% for Senator Flávio Bolsonaro, according to the Nexus/BTG poll released on Monday (21). With a margin of error of two percentage points, the three-point gap counts as a technical tie. Writer Augusto Cury ranks third with 6%.

The survey was commissioned by Banco BTG Pactual, a Brazilian investment bank, and polled 2,006 voters nationwide by phone between Friday (18) and Sunday (20). The margin of error is two percentage points in either direction, with a 95% confidence level, CNN Brasil reports. The poll is registered with the TSE, Brazil's top electoral court, under code BR-00485/2026.

First round: a stable race

Behind Cury come Ronaldo Caiado (PSD) with 5% and Renan Santos (Missão) with 3%. Former Minas Gerais governor Romeu Zema (Novo) and Samara Martins (UP) have 1% each, and the remaining presidential candidates scored zero. Blank, null or no-candidate answers totaled 4%, while 2% were undecided or gave no answer.

Compared with the previous wave, Lula slipped from 42% to 40% while Flávio held at 37%, a shift within the margin of error. Flávio Bolsonaro is the son of former president Jair Bolsonaro, who was convicted and jailed for attempting a coup, as Folha de S.Paulo notes. Leonardo Avalanche, announced by the PRTB on Tuesday (15) to replace Pablo Marçal, whose candidacy the TSE rejected unanimously, did not register in the poll.

Runoff and other matchups

In a potential runoff between the two leaders, Lula has 46% and Flávio 45%, a technical tie like the previous week's 47% to 46%. The poll also tested other matchups, according to bahia.ba: Lula ties Caiado at 45% to 45% and edges Cury 44% to 43%, both within the margin, and leads Zema 47% to 41% and Renan Santos 47% to 39%, gaps beyond the margin of error.

What the data confirm is a snapshot of voter intentions between September 18 and 20: the two front-runners technically tied in both rounds, with weekly movement inside the margin of error. What they cannot show is the election outcome, with the first round about two weeks away. To follow polls responsibly, readers can check this survey's registration in the TSE electoral research portal using code BR-00485/2026, compare weekly waves rather than a single reading, and treat any gap of two points or less as a tie, which this poll cannot distinguish.