Brazil’s 2018 election began as a political crisis and ended as a mystery: how did an app for private messaging, where grandmothers forward recipes and good-morning cards, turn into the most powerful and least understood campaign machine in Latin American history?
To understand what happened in Brazil in the fall of 2018, you first have to grasp the scale of a single figure: more than 120 million Brazilians used WhatsApp—out of a population of roughly 210 million. More than the number of people watching the country’s major TV channels. More than the combined readership of every newspaper. The messenger, then owned by Facebook Inc., had long since ceased to be merely a means of communication: people used it to book doctor’s appointments, call a plumber, arrange supplies for small businesses. Brazil is one of those countries where WhatsApp effectively replaced SMS, email and, for a substantial part of the population, the rest of the internet altogether.
Groups—up to 256 members each—formed around anything that bound people together: family, church parish, parent-teacher association, condominium neighbors, work colleagues. A person could belong to ten, twenty, thirty groups at once. A message sent to one group could be forwarded to another with a single tap. From there—to a third. Chains of forwards worked like a capillary network: slower than a viral tweet, but deeper, because every message came from someone familiar—aunts, pastors, bosses. End-to-end encryption meant that neither WhatsApp itself, nor researchers, nor journalists, nor electoral authorities could peer into these conversations unless one of the participants chose to share their contents voluntarily. Political communication had moved into a space that was, by definition, a blind spot for outside observers.
On September 6, 2018, in the city of Juiz de Fora, PSL candidate Jair Bolsonaro was stabbed during a campaign rally. He spent around a month in hospital. For any other candidate, that would have meant disaster: no debates, no travel, no face-to-face meetings with voters. Bolsonaro turned it into an advantage—or, more precisely, his supporters did. The campaign moved almost entirely into the digital realm, and WhatsApp became its main artery.
According to Datafolha (a poll published in early October 2018, on the eve of the first round), 61% of Bolsonaro’s voters named WhatsApp as their source of political news, compared with just 38% of voters for his opponent, Fernando Haddad of the PT. A gap of twenty-three percentage points is not statistical noise. It is two different information worlds. Bolsonaro’s supporters built a broad, organic digital base that existed and operated long before journalists began writing about the “WhatsApp election.” They created groups, coordinated forwards, produced content—memes, videos, voice messages—with an energy that is hard to explain through outside organization alone.
One detail here strikes me: a candidate physically confined to a hospital bed became omnipresent—in every voter’s pocket, in every family chat. Traditional campaign logic, built around televised debates and street rallies, ran into something for which it had neither the tools of analysis nor the language to describe.
It would be a mistake to tell this story as though a single messenger had overturned the politics of a country of two hundred million people. By 2018, Brazil had gone through a series of shocks, each one eroding trust in existing institutions. Operation Lava Jato—a vast anti-corruption investigation—exposed a kickback system linking the biggest construction companies with politicians from different parties. Former president Lula da Silva of the PT was convicted and jailed. President Dilma Rousseff, also of the PT, went through impeachment in 2016. The economy stagnated. The country’s level of violence remained among the highest in the world.
Anti-PT sentiment—a deep, emotional weariness with the party that had ruled for thirteen years—was not the product of WhatsApp forwards, but the result of millions of people’s real political experience. Distrust of traditional media—newspapers and TV channels seen as part of the establishment—also existed long before the election. WhatsApp did not create these feelings. It gave them infrastructure: fast, emotional, shielded from outside scrutiny. Put crudely, the messenger was not the source of the fire, but the wind that carried the sparks.
On October 18, 2018—ten days before the second round—the newspaper Folha de S.Paulo published an investigation by journalist Patrícia Campos Mello. The report claimed that a number of companies had paid for contracts for mass WhatsApp messaging (disparos em massa) directed against the PT and in support of Bolsonaro. Those named included the Havan retail chain and its owner, Luciano Hang. The contracts, according to the newspaper, were worth as much as 12 million reais.
The mechanism was this: specialist agencies—Quickmobile, Yacows, Croc Services, SMS Market, Deep Marketing, Kiplix—offered mass-messaging services. Prices varied: 0.08–0.12 reais per message if the candidate’s own contact database was used, and 0.30–0.40 reais if the agency provided its own database. The practice was illegal on two grounds: Brazilian law prohibited corporate donations to election campaigns, and the use of third-party databases violated rules governing the handling of personal information.
Bolsonaro’s campaign officially declared only 115,000 reais paid to the agency AM4 Brasil Inteligência Digital. The gap between that sum and the scale described in the investigation became the central question: was there direct coordination between the campaign and the companies funding the mass messages? Those implicated denied the allegations. Bolsonaro called the publication an attack by hostile media.
On December 2, 2018, Folha published a follow-up: a former Kiplix employee named Hans River do Rio Nascimento described a fraudulent scheme in which the names and CPF numbers (Brazil’s equivalent of a taxpayer ID) of 10,000 elderly people—born between 1932 and 1953—were used to register SIM cards in so-called chip farms (devices that make it possible to manage large numbers of SIM cards and phone numbers simultaneously for automated messaging). The elderly people whose data had been stolen most likely had no idea that dozens of phone lines had been registered in their names, sending out political messages.
Following the publications, WhatsApp blocked the accounts of the agencies mentioned. But the legal outcome proved far less dramatic than the journalistic investigations. On February 9, 2021, Brazil’s Superior Electoral Court (TSE) dismissed two lawsuits brought by the Brasil Soberano coalition (a bloc of the PDT and Avante), which had demanded that the election results be annulled over the mass messaging. No reliable evidence of a direct link between Bolsonaro’s campaign and the messaging schemes was presented—on the merits, the judges were unanimous; the only split, six votes to one (Edson Fachin’s position), concerned the procedural question of joining the case to other, still unresolved lawsuits. That did not mean the mass messaging did not exist—their existence had been documented. The court found the specific legal element unproven: coordination between the campaign and those paying for the messages.
While political messages spread through millions of chats, a small coalition of Brazilian fact-checkers was trying to build a dam with bare hands. The Comprova project, bringing together journalists from different newsrooms, received almost 47,200 requests to verify information through WhatsApp. The team investigated up to six stories a day—and that was catastrophically little.
One detail shows the scale of the problem better than any statistic: when a fake voice message about the attack began spreading through WhatsApp groups after Bolsonaro was stabbed in September, it took Comprova three days to debunk it. In three days, the audio had time to pass through hundreds of thousands, perhaps millions, of groups. The correction, even once it appeared, moved more slowly—because the lie was emotional, shocking, “hot,” while the debunking was boring and demanded attention.
The #CheckBR coalition—Agência Lupa, Aos Fatos, Comprova, Boatos.org, e-Farsas, Fato ou Fake—worked together with the TSE, which opened its doors to fact-checkers on October 22, 2018. In the 48 hours before the second round, they checked 50 false messages. Fifty sounds impressive, until you remember that in a country with 120 million WhatsApp users, thousands upon thousands of unreliable items were in circulation: fabricated stories about political opponents, conspiracy tales about schools supposedly “corrupting” children, panic messages about crime, fake “inside information” about election fraud.
I think this is the asymmetry that determined everything: the production of false content was distributed, cheap, and fast, while checking it was centralized, labor-intensive, and slow. The fact-checkers worked like surgeons trying to operate on a patient being bitten by thousands of mosquitoes at once.
WhatsApp began responding during the campaign itself, blocking accounts tied to automated mass messaging and introducing labels for forwarded messages—a small arrow indicating that a text or image had not been written by the sender, but received from somewhere else. The measure was cosmetic: most users paid no attention to the arrow, and even if they did, it did not affect their trust in the content. A message from an aunt in a church group remained a message from an aunt in a church group, regardless of the icon.
A more substantial step came after the election. In January 2019, WhatsApp lowered the forwarding limit from 20 to 5—meaning that a user could now forward a particular message to no more than five chats at a time, rather than twenty. Brazil’s electoral crisis became one of the main arguments for this decision, though similar problems had emerged in other countries—in India, for instance, viral WhatsApp messages had led to outbreaks of mob lynching.
Did the restriction help? A study published on arXiv (by Melo, Garimella and co-authors, number 1909.08740), which analyzed 6 million public messages in Brazil, India, and Indonesia, showed that the new limit slowed the spread of content but did not stop high-reach viral messages. Those who wanted to spread something widely simply did it in several rounds—or used multiple accounts. The patch reduced the pressure of the water, but it did not fix the pipe.
On October 28, 2018, Jair Bolsonaro won the second round with 55.13% of the vote against Fernando Haddad. The victory was decisive—a margin of more than ten percentage points left no room for arguments about the result’s legitimacy in an arithmetic sense. But arguments over how that victory was achieved continued for years.
In these arguments, it is easy to fall into one of two extremes. The first: WhatsApp “stole” the election, mass messaging brainwashed millions, the result is illegitimate. The second: the messenger had no influence at all, Bolsonaro won because of real public sentiment, and journalists inflated the story. Both extremes are convenient, and both are wrong.
Reality lies somewhere in between. Bolsonaro had powerful organic support, born of a years-long crisis of trust in the political class, fear of crime, and the conservative turn of a significant part of Brazilian society. That support was real. But WhatsApp did something no other platform could: it turned political campaigning into a personal message. In a space where there is no difference between a friend’s movie recommendation, a party slogan, commercial spam from a paid agency, and a fabricated story about an opposing candidate—in that space, the very category of an “information source” loses its meaning. Everything comes through one window. Everything looks the same. Everything comes from someone you know.
Brazil’s 2018 election became the first major case in which a private, fragmented, technically opaque communications network became the arena for mass political mobilization on the scale of an entire country. Subsequent elections in other states—and Brazil’s 2022 election—took place with this precedent in mind, with new rules, with attempts at regulation. But the underlying architectural problem has not gone away: encryption, which protects citizens’ private lives, also makes coordinated manipulation invisible, and no technical solution has yet managed to resolve that contradiction without destroying one of those two values.