The blue button in the corner of the screen opens the translation into Libras (Brazilian Sign Language) (VLibras).
Runoff, made simple
Main question
—
Average of the most recent polls from — pollsters, counting only votes given to one of the two. Flávio is ahead in — of them; Lula, in —.
Valid votes are the votes given to one of the candidates, not counting blank and null votes. This is how the TSE decides who wins.
Margin of error is how much a poll's result can vary up or down, because it only hears from a portion of voters.
Statistical tie happens when the difference between the two is smaller than that margin: you cannot say who is ahead.
—
Among the — pollsters that have run more than one poll in the runoff, the average gap was — points in favor of — in the first round of polls and is — points in favor of — in the most recent. The average of all pollsters is in the Polls tab.
—
Flávio got —% e Lula, —%, on October 4. Since no one got more than half, the two compete again.
The share of each person's checked statements that was classified as false, by fact-checker. Each one chooses what to check and uses its own criteria, which is why the numbers vary so much.
—
From — measures collected, Lula is ahead in — (for example, —) and Flávio in — (for example, —). Each measure tells part of the story, and none shows who will win.
— debunked rumors
There are rumors against both candidates (— against Lula and — against Flávio). The count includes only what fact-checkers and the TSE have already debunked, not everything in circulation. Before sharing anything, check the source.
—
From — to —, Brasília time. —
| Day | Lula | Flávio |
|---|
Debates and interviews, moment by moment
The most recent appears first. Moments and fact-checks come from live coverage by news outlets and fact-checkers, with a link to each. A statement only appears as checked when the fact-checker has published the seal. During the debate, this page receives updates roughly every half hour.
Who is ahead, poll by poll
Here are the national polls for the runoff. The chart shows all rounds; the table shows the latest from each pollster. On the chart, each dot is a poll: the higher above the middle line, the further ahead Flávio is; the lower below it, the further ahead Lula is.
On average across each pollster's latest round, Flávio has —% and Lula —% of valid votes, a difference of — point. Flávio is ahead in — of — pollsters; Lula, in —; tie, in —.
Each dot is a poll, and its height shows points of lead the leader has in valid votes, not each one's percentage. Example: the Vox poll of 18/10 gave Lula 51% and Flávio 49%, so the point sits at L +2. Above the tie line, Flávio is ahead; below it, Lula. The gray lines connect polls from the same institute, with its name in the right-hand column. Tap a point to see the numbers.
Valid votes; the bar shows which way the difference leans. The pollster's name links to the original report.
The polling institute Veritá, which appears in this list, has no relationship with Clareia Eleições (formerly called Verita Eleições). The names are similar, but they are different initiatives.
| Pollster | Date | Flávio | Lula | Difference | Sample | Margin | Source |
|---|
% of valid votes, October 4 (TSE).
Voting intention in the runoff among voters of defeated candidates. The rest say blank, null or undecided.
| Group | Poll | Flávio | Lula |
|---|
How each one shows up on social media
How the two candidates are doing on social media. Remember: likes and followers show who is mobilizing supporters, not who will win.
Words in this tab: interaction (or engagement) is each like, comment or share; mention is each time the candidate's name appears in a post; view is each time a video is viewed.
The candidates' official profiles, according to Bites, Quaest, TV Globo, Projeto Brief and AP Exata (the source for each row is alongside). Each metric on its own scale.
Readings from Quaest and Bites on the digital environment since the first round.
| Date | Signal | Source |
|---|
Endorsements, schedule and Electoral Justice
What each candidate will do in the coming days and what has already happened in the campaign, from newest to oldest.
Commitments announced by the campaigns; they may change.
| Day | Lula | Flávio |
|---|
Endorsements, TSE decisions, debates and strategy changes, from newest to oldest.
| Date | Type | What happened | Source |
|---|
What is true, what is a rumor
Brings together fact-checks of statements by both candidates made by outlets such as Lupa, Aos Fatos, Estadão Verifica, g1 Fato ou Fake and Ranking dos Políticos. Each one uses its own labels and criteria, grouped here into three bands: true, imprecise or missing context, and false. The scores only show the statements each fact-checker chose to check, cannot be compared between fact-checkers and do not measure the whole campaign. In some cases fact-checkers even record what was said differently. Tap the fact-checker's name to open the original fact-check.
Only statements by the candidates themselves, in interviews, livestreams and debates, adding up all fact-checkers. The result varies a lot from one fact-checker to another: see the table just below.
of each one's checked statements were classified as false: — of — for Flávio and — of — for Lula.
Percentage of false statements, adding up all fact-checks up to each date.
The result changes depending on the fact-checker. Each one chooses which statements to check and uses its own criteria:
| Fact-checker | Candidate | False | Checked | False rate |
|---|
Counts only statements by the candidates themselves in interviews, livestreams and debates, not third-party rumors. Since fact-checkers do not check everything each one says, the rate is a snapshot of the checked statements, not of the whole campaign.
Claims in each band, by fact-checker and interview.
| Candidate | Fact-checker | Event | Date | True | Imprecise | False | Total | Note |
|---|
Each fact-checked claim was classified by the panel according to who it targets: attack on the opponent, defense of oneself or one's own government, or proposal and other topics. It counts the statements listed one by one below; that is why the total is lower than the fact-checkers' scoreboard, which adds up everything each fact-checker counted.
| Who said it | Type of statement | True | Imprecise | False | Total |
|---|
Candidates' statements in interviews, livestreams and debates, with the fact-checker's original seal and a short explanation. These are the statements we could list one by one; the scoreboard above is larger because it adds up the counts published by each fact-checker.
| Date | Who | Claim | Type of statement | Seal | Why | Fact-checker |
|---|
Number of debunked rumors by origin and target. The origin is only counted when the fact-check or the TSE decision says who published it; in most cases it is not identified.
| Origin | Against Lula | Against Flávio |
|---|---|---|
| Flávio's camp | — | — |
| Lula's camp | — | — |
| Not identified | — | — |
The table counts only rumors already debunked by fact-checkers or the TSE, not everything that circulates on social media.
Content that circulated against the candidates and that fact-checkers debunked or the TSE ordered removed. Each card says who it targeted, what is known and who debunked it, with the link.
| Date | Against | Favors | Origin | Rumor | Topic | Seal | About the origin | What is known | Who debunked it |
|---|
Received a screenshot? Check here
Each piece of information here has a code, which appears like this: check 3F9A-12C7. Enter below the code shown in the screenshot and see how the information actually appears in the page data. If the screenshot text is different, it was altered.
Twice a day we check the published information against the original sources. When we find an error, we correct it and record here what changed and why.
No corrections so far.
| Date | What was corrected | Before | After | Reason | Source |
|---|
The code is calculated from the exact content of the information. If someone changes one letter or number, the code no longer matches.
The data lives in a public repository on GitHub, which keeps a timestamped record of everything that was published and corrected. When a piece of information is corrected, the old code remains here, marked as an old version, so that an old, genuine screenshot can also be checked.
Screenshots without a code, or with a code that does not appear here, did not come from this page.
So no one votes in the dark
New name: Verita Eleições is now called Clareia Eleições. We changed it to avoid confusion with the Instituto Veritá, which does election polling and with which we never had any relationship. The purpose and the way we work remain the same. Clareia does not conduct polls: it only gathers those that other institutes have published, including Veritá's, with a link to the source.
Clareia gathers in one place what has already been published about the elections in Brazil, with the source for each piece of information and both sides held to the same standard, so you can understand and decide for yourself.
The search, summaries and organization are done with artificial intelligence, with automatic checks twice a day; Instagram posts are approved by a person before they go out. Found an error? Contact us on Instagram @clareiaeleicoes or open an issue on GitHub.
How much of what is on the site is about each candidate, counted automatically from the data. It lets anyone check whether Clareia gives both the same space.
| What we count | Flávio | Lula |
|---|
The numbers do not need to be identical: they depend on what pollsters, fact-checkers and the press have published. Large differences appear here for everyone to see.
This site is an information aggregator. It does not conduct its own polls or fact-checks: it gathers what polling institutes, news outlets and fact-checking agencies have already published, and always shows the name and link of the original source, where the full content is.
To respect the copyright of whoever produced each piece of content, only short summaries written in our own words and brief excerpts of the candidates' public statements, in quotation marks, appear here. We do not copy whole reports or fact-checks, nor photos, videos, charts or logos from other sites. The poll numbers and the fact-check seals belong to whoever published them. This site has no ties to any of these outlets, institutes or fact-checkers, nor to the campaigns.
If you are the author of any cited content and want a correction or removal, open an issue on GitHub.
Made with artificial intelligence. The collection of news, the summaries, the organization of the data and the code of this page are done with the help of AI (Claude, by Anthropic), with automatic data checks and quality review twice a day. Even so, there may be errors: always check the source and see the corrections in the Verify tab.
Found an error, want to suggest a fact-check, or are the author of cited content? Message us on Instagram @clareiaeleicoes or open an issue on GitHub. Every correction is recorded under .
The site's data is open and can be downloaded at github.com/aluadocoelho/Eleicoes-de-2026.
This page works with a screen reader (TalkBack, VoiceOver, NVDA) and by keyboard: use Tab to move around and the arrow keys to switch tabs. All charts have a text description. The button A+ at the top increases the font size, and the moon button switches to dark mode. The blue VLibras button, in the corner of the screen, translates the content into Libras. The page has no audio or video: everything is written.
—·—
Clareia Eleições (formerly Verita Eleições) is an independent site, with no ties to parties, campaigns or polling institutes. We have no relationship with Instituto Veritá, an election polling firm. This site was made with artificial intelligence. Information aggregator: the original content belongs to the cited sources, with a link on each item. Code and data: github.com/aluadocoelho/Eleicoes-de-2026.
Data gathered from published reports and fact-checks, with a link to each source, updated every hour, from 7h to 23h, and more often during debates. The data is open and can be downloaded at github.com/aluadocoelho/Eleicoes-de-2026 (data under CC BY 4,0; the numbers and seals belong to the institutes and fact-checkers cited). The page counts visits with GoatCounter, which does not use cookies or store personal data. If you accept, we also use Google Analytics, which uses cookies, to understand which tabs are read most; it stays off if you decline.
If you accept, Google Analytics uses cookies to count visits and see which tabs are read most. We do not use it for ads. If you decline, the site works the same.
We let you know when a new poll, fact-checked statement or debunked rumor comes out. It works while this page is open, even in another tab.