Of the 190 Brazilian politics channels on YouTube monitored by Botponto, 20 mentioned Polymarket between April 24 and May 27, 2026. Even though the platform has been banned in Brazil for a month, Núcleo found that 44 videos used it to gauge potential voting intentions, even though it isn't an electoral poll.

🤖
Botponto is a tool that scrapes and analyzes YouTube videos, tracking keyword mentions and identifying discourse trends. It was developed by Núcleo with support from the International Center for Journalists (ICFJ).

In most cases, these are right-wing and far-right influencers and journalists discussing the electoral landscape around the prospective presidential candidates, chiefly Lula (PT) and Flávio Bolsonaro (PL).

The videos average 32,000 views. The most-watched is a video published on April 25, 2026 by the channel Te Atualizei — with 401,000 views as of May 29, 2026 — in which influencer Bárbara Destefani states that "Polymarket is a benchmark for election observation worldwide."

The quotes and video titles in this visualization were originally in Portuguese. They have been translated to English for publication.

Sites like Polymarket and Kalshi serve as intermediaries for users betting on the outcome of real-world events — from political decisions, economic indicators, and sporting matches to cultural awards and even wartime events.

For lack of specific regulation, Polymarket and 26 other prediction-market platforms, or "betting exchanges," were banned in Brazil on April 24, 2026.

The National Monetary Council (CMN) issued a resolution banning the offering and trading in Brazil of derivative contracts whose underlying assets are tied to entertainment, political, or sports events. This means they're no longer treated as financial investments but as a form of unauthorized betting that, according to the Ministry of Finance, must follow the rules of the Secretariat of Betting and Prizes (SPA).

What are derivative contracts?

According to the Securities and Exchange Commission (CVM), derivative contracts are "financial assets whose value results (or derives — hence the name), in whole or in part, from the value of another financial asset (the underlying asset) traded on the spot market. These contracts are traded on an exchange (a stock exchange) or directly with a financial institution (over-the-counter)."

In other words, the value of the investment can vary according to the dollar exchange rate, interest rates, stock-market indices, equities, and commodities (such as coffee, gold, and so on). These contracts can be used for various purposes, among them so-called "hedging," when the goal is to protect the value of an asset position (an investment portfolio) or a liability position (debts) against possible future fluctuations in a given asset or liability, for example. The CMN resolution did not affect these contracts dealing with economic and financial indicators.

The SPA argues that the contracts these platforms offer replicate the betting-exchange model described in Ordinance 1.231/2024: bettors wager against one another, and the odds are set between them rather than by the operator (the betting houses), which can take a commission on the net winnings. On top of that, the 2018 law that legalized sports betting — Law No. 13,756 — only allows fixed-odds betting on sporting events.

In the technical note that underpinned the ban, the SPA also notes that "operating prediction markets not only falls short of any claim to legality from a regulatory standpoint, but may also directly violate public-order rules designed to protect the integrity of the democratic process."

The passage refers to Resolution No. 23.735/2024 of the Superior Electoral Court (TSE), which bars the use of digital platforms that offer or promise — whether for free or for a fee — "goods, products, or advertising linked to candidates or to the outcome of the election." Violations can constitute an electoral offense, such as abuse of economic power or illegal vote-buying.

The SPA's press office told Núcleo that "the blocking order has already been implemented by most providers" following notification from Anatel, the telecommunications regulator.

If it's banned, why is it still being used?

One explanation for why these platforms have caught on is that news outlets in the United States have partnered with these sites — which also raises questions about journalism's role in spreading this kind of betting. In the US, these sites are regulated.

CNBC, CNN, Fox Corporation, and the Associated Press have partnered with Kalshi to bring prediction-market data into their coverage of both economic and political topics. And Dow Jones, the publisher of the Wall Street Journal, has done the same with Polymarket.

"The right has always been much more open to the market — both on public policy and in its belief that the market is the best way to understand society," says David Nemer, a technology anthropologist and professor at the University of Virginia.

"Given that mindset, and the fact that the United States has always been a model society — especially for people on the right — and that using [Polymarket] there has become common practice, the right tries to replicate that model in Brazil," he explains.

Is it a crime to use a VPN to access Polymarket?

Some of the influencers and journalists identified in the videos monitored by Núcleo mention — and even recommend — that, although Polymarket is banned in Brazil, it can still be accessed via VPN: a virtual private network that masks the user's location.

In this case, the reporters didn't need to use one. The fact that Núcleo's reporters are based in different states, for instance, made it possible to confirm that Anatel's blocking order had not yet reached every provider. In São Paulo, access was blocked; in Goiás, as of Wednesday, May 27, it still wasn't. In Rio de Janeiro, where the editor is based, the platforms remained accessible at least until Friday, May 29.

Experts consulted by Núcleo point out that nothing in the Criminal Code criminalizes the use of a VPN on its own.

"It's important to keep in mind that a VPN works very much as a legitimate tool for accessing information — a security tool — so the federal government itself has guidelines for civil servants and officials to use them, and many companies do the same," explains Thobias Prado Moura, director of Internet Society (ISOC) Brazil.

"The text of the resolution, the notes from the CVM and the Central Bank, and the positions taken by the Ministry of Finance are very much aimed at operators, not users," stresses André Fernandes, director of the Recife Institute for Research on Law and Technology (IP.rec). "So if operators systematically offer users access to the platform, encouraging the use of VPNs, they would be held liable — but I don't see the user's behavior as constituting a criminal offense".

One example of this debate is the 2024 blocking of X. Justice Alexandre de Moraes, of the Supreme Federal Court (STF), went so far as to order fines for anyone using a VPN to access the social network, but backtracked the same day "to avoid disruption for other companies."

Paulo Rená — who holds a doctorate in law from the University of Brasília (UnB), is a researcher at the Reference Center on Internet and Society (IRIS), and sits on Núcleo's advisory board — stresses that any illegality, when it exists, lies in the act of accessing certain content, not in the tool used to reach it. "Things that are banned are banned regardless of the means of access," he says, comparing the situation to buying a prohibited item, which is illegal whether the purchase is made in a domestic or a foreign store.

He adds a technical complication: people who access a blocked site via VPN don't always know they're doing so. Apps like Telegram run on virtual private networks built into the app itself, so a user can reach a banned channel without being aware of either the ban or the fact that they're using a VPN.

Veridiana Alimonti, a director at the Electronic Frontier Foundation (EFF) who holds a doctorate in law, echoes the point. She notes that nothing in Brazilian law or case law provides a basis for treating VPN use as an offense separate from the access itself, and she warns against what she calls "a trap": creating mechanisms to judge the legitimacy of VPN use case by case, based on what the person accessed.

By that logic, she says, any digital security tool — including end-to-end encryption — could end up being judged by what it shields. "Using a VPN or any other digital security tool should not, in itself, be penalized beyond the criminal conduct," she adds.

Thobias Moura also points out that this is a complex debate, and that ways need to be found to enforce court and administrative decisions without undermining the information-security and privacy principles that VPN platforms provide.

"A VPN provider is protected by the same constitutional guarantees — confidentiality of data communications, and protection of privacy, intimacy, and private life. By its very technical design, a VPN provider has no access to what passes through it. So for a provider to gain access and know, say, that someone is accessing Polymarket, it would have to use techniques that would necessarily undermine that very idea of privacy and freedom," he explains.

Núcleo asked the Ministry of Finance why prediction-market sites are still available to some users in Brazil and whether there was any timeline for when all of them would be blocked. The press office said that "the block has already been implemented by most providers" and that the offering and promotion of these platforms "remain prohibited."

In Brazil, for example, one of the videos identified by Núcleo — posted on May 21, 2026 by the channel Missão Avança — reproduced a post on X by Guga Chacra, GloboNews's New York-based international politics commentator, in which the journalist noted that "the probability of victory" for prospective candidate Renan dos Santos (Missão) "is quite significant," based on a look at the presidential-election betting on Polymarket.

Screenshot of a post on X by Guga Chacra, GloboNews's New York-based commentator, displaying Polymarket's prediction-market chart for the "Brazil Presidential Election." He wrote "It's striking how Renan Santos has been climbing in the betting markets for Brazil's presidential election. His probability of winning is quite significant". The platform put Lula in the lead at 47%, followed by Flávio Bolsonaro (24.3%), Renan Santos (14%), and Romeu Zema (5.5%). (Reproduction/X)

Months earlier, in December 2025, Fantástico, a Sunday news show on Rede Globo, ran an extensive report interviewing Kalshi co-founder Luana Lopes Lara — who had joined Forbes's billionaires list — without scrutinizing the issues surrounding prediction markets.

Money vs. polls

In the videos monitored by Núcleo, one of the main arguments influencers use to vouch for Polymarket's reliability is that people put real money into the platform, meaning they have skin in the game.

Leandro Maciel, a finance professor at the University of São Paulo (FEA-USP) and a researcher in financial econometrics, explains that, unlike a voter-intention poll, prediction markets can be manipulated and lack any methodological concern for sampling bias:

"If certain groups want to reinforce a particular view, expectation, or political stance, they can use platforms like this to prop up certain narratives," he explains. "The caveat is that the results seen in these markets may contain sampling bias — that is, they may reflect only the profile of the people who typically access and trade on these platforms."

He also points out that those who put money into these platforms are looking to make a profit.

“So the focus is on getting the outcome right, not necessarily on expressing a preference for a specific candidate. That differs from a voter-intention poll, in which the respondent, in theory, has no financial motivation when answering. In that context, the answer tends to reflect their actual voting intention.”

According to an analysis by the Pew Research Center, global trading volume on Polymarket and Kalshi — the two largest prediction markets — rose from less than $5 billion in September 2025 to around $24 billion in April 2026. The researchers note that this is higher than the average monthly amount wagered on legal betting sites in the United States, which came to around $14 billion.

Sports is still the dominant category on both platforms — 80% of total trading volume on Kalshi and 39% on Polymarket — according to the study. It is followed by cryptocurrencies (7% of total volume on Kalshi, 20% on Polymarket) and politics (4% on Kalshi, 32% on Polymarket).

In Polymarket's case, transactions are made in cryptocurrency, which makes it impossible to trace where the bets come from. The transactions are publicly recorded on the blockchain, where you can see how many wallets placed bets and the amounts involved — but since the platform doesn't require users to identify themselves, there's no way to know who is betting: whether they're Brazilian or foreign, a voter or a bot, or whether a single person is running multiple accounts.

💡
Blockchain is a distributed, immutable digital ledger technology — like a database, but public and decentralized: it isn't stored on a central server but maintained simultaneously across thousands of computers around the world. It's the technology behind cryptocurrencies like Bitcoin and Ether, and the one that records the bets placed on Polymarket.

Signs of manipulation and insider trading

A New York Times story published in May 2026 identified more than 80 Polymarket users whose bets showed signs of insider trading — when a bettor places a large sum on an outcome because they have access to information the public doesn't. According to the newspaper, these users turned a profit across nearly 30 different topics since 2024, most of them tied to military operations.

A report by the Anti-Corruption Data Collective (ACDC) examined all 435,672 bets settled on the platform between January 2021 and March 2026 — a total of $54.4 billion. It concluded that certain categories of political markets, especially those tied to military actions, show systemic signs of insider trading.

The study shows that political markets made up just 4% of the platform's bets but still accounted for more than a third (36%) of all the money wagered — $19.7 billion.

The report's authors note that a higher-than-expected success rate doesn't, on its own, prove the use of inside information, but the pattern shows up across a range of indicators: beyond the high success rate, there is a cluster of wallets with near-perfect win rates and a string of winning bets placed in the final hours before a market closes. Taken together, these signals point to a structural risk.

Another point the researchers raise is that the blockchain's transparency, combined with the lack of identity verification, creates an amplifying effect: since anyone can follow the bets in real time, high-frequency traders can spot suspicious moves and copy them, further widening the information advantage over the ordinary bettor.

One of the researchers' recommendations is precisely that Polymarket start requiring identification from all bettors, as its competitor Kalshi already does.

Unlike prediction markets, electoral polls interview people and follow a scientific methodology. In Brazil, every public opinion poll on elections or candidates must be registered with the electoral courts, where anyone can look up the results, the questions asked, and how the poll was conducted.

One of the main concerns in opinion polling is designing a diverse sample that can represent a cross-section of the population, explains Neale El-Dash, who holds a doctorate in statistics from the University of São Paulo (USP) and founded PollingData, a website that aggregates electoral polls. 

"If you don't think about how to interview these people, about the logistics of data collection, you end up leaving subgroups of the population with no chance to take part in your survey, and you'll have segments that go unrepresented," he points out.

That doesn't mean public opinion polls can't be looked at critically. The researcher stresses that several factors can affect the outcome — the way interviews are conducted (in person, by phone, or online), the type of questionnaire used, and how much access interviewers have to different segments of the population — since every methodology has its advantages, disadvantages, and limitations. What matters, he says, is that all of this be transparent to the public.

Credibility and disinformation

Heloísa Massaro, director of research and operations at InternetLab, stresses that we live in a climate of widespread distrust, in which people feel lost amid the sheer volume of information. "People end up building what are almost their own information ecosystems," she says.

For David Nemer, it's also ethically questionable for news outlets to use prediction-market platforms in their coverage. "I think using this in the press is a major betrayal of the principles of journalism, because you're no longer reporting whether something is true or not. You're reporting a fact based on the probability of its being a fact," he says.

How we did this

After Polymarket was banned in Brazil — and following the post on X by GloboNews commentator Guga Chacra — we used BotPonto to find out what was being said about the platform on YouTube, and whether anyone else was using prediction-market tools as "election polling firms".

Within the tool, under the "Open Search" tab, we ran a free-text search for "Polymarket." The search returned 53 mentions of the term, each entry listing the channel, the video title, a link to the exact moment it was mentioned, a transcript of what was said, the video's date, the view count, and the channel's political leaning. We downloaded all this data as a .csv file and, working from a spreadsheet, watched the videos, analyzed the content, and removed any records from before April 24, 2026, so as to consider only mentions that occurred after the ban.

ℹ️
This story was originally published in Portuguese and translated with the help of artificial intelligence, with close review by human editors, according to Núcleo's AI use policy.
Reporting Sofia Costa, Jeniffer Mendonça
Artwork Rodolfo Almeida
Editing Alexandre Orrico
Translation Sofia Costa