Anonymous 'independent expert report' alleging fraud in Brazil's runoff is contradicted by electoral court data

A document titled "Relatório Pericial Independente" (Independent Expert Report) has been circulating on social media, alleging fraud in Brazil's presidential runoff. It carries no author's name, and the press has labelled it "apócrifo", meaning anonymous or of unknown authorship. Fact-checkers and mainstream outlets say it distorts public data from the electoral court (TSE).

What's being claimed

According to Estadão Verifica, the report says it analyses TSE microdata and makes several points:

• Flávio Bolsonaro (PL) would have won if you count only the voting machines closed by 5 p.m.
• Votes cast after 5 p.m. can be assigned to one candidate.
• The TSE's figure of roughly 34,000 late votes is false.
• 8.6 million voters voted without biometric identification, described as the process's "biggest material vulnerability".
• The microdata were released only 72 hours after the result was proclaimed.

The document was shared by supporters of Senator Flávio Bolsonaro and by U.S. commentator Joey Mannarino, a Trump sympathiser, whose Instagram post reportedly drew over 400,000 likes.

What the evidence shows

The main claim fails on arithmetic. Lula won with about 64.2 million votes to 59.8 million, a gap of roughly 4.4 million. Only about 34,000 votes were recorded after 5 p.m.: 32,399 between 5 and 6 p.m., 1,981 between 6 and 7 p.m., and 13 after 7 p.m. Even if every one of them went to a single candidate, which no one can know, the result would not change. Researchers from USP's Monitor do Debate Político told Estadão that most of these votes came in the first ten minutes, which fits people still waiting in line. Late votes are also not new: 31,714 people voted after closing time in the 2022 runoff.

The report's attempt to link late votes to one candidate also collides with ballot secrecy. The TSE says the Digital Vote Record keeps choices without tying them to voters' identities.

On biometrics, the TSE did not confirm the 8.6 million figure. Biometric registration is not mandatory, and about 141.3 million of 158.7 million voters (89%) had it. When a fingerprint is not recognised, a fallback applies: the voter types their birth year, signs the voting register, and the poll worker who authorised the vote is identified. The published files do not say whom those voters chose.

The "72 hours" claim is also contradicted. The electoral court says bulletins were printed and posted at polling places, then published online, and the vote record and machine logs are available for consultation.

One caveat: our evidence comes mainly from the electoral court's statements as relayed by Estadão Verifica, not from our own analysis of the microdata.

How outlets covered it

Four outlets ran the story. Aos Fatos said the report "distorts" TSE microdata to allege fraud. Estadão Verifica published the detailed point-by-point check. Estadão and Terra carry the same headline, calling the fraud allegation "falsa". For these three we only had headlines, so we cannot describe their content. None of the provided sources defends the report, and there is no coverage from outlets that endorse it beyond the social media shares described above.

What we still don't know

• Who wrote the report. No source identifies the author.
• Where the 8.6 million figure comes from, and whether it can be reproduced from public data. The TSE did not confirm it.
• Whether independent analysts will publish their own breakdowns of the microdata. Estadão cites several outlets and USP researchers reaching similar conclusions, but we did not see their full methods.
• How far the claims will keep spreading, given the large reach of the shared posts.