In 30 seconds
8 seconds of TV. With a small party and no debates after the stabbing, Bolsonaro won the 2018 runoff with 55.1% of valid votes.
A pyramid of WhatsApp groups: few producers, many volunteer admins and millions of ordinary people. On top of that came mass messaging paid for by companies, according to Folha.
Distribution by ordinary people made up for missing TV time. Bought lists, hidden money and lies brought lawsuits and harm no democratic campaign should repeat.
How to read this case
There is no toolkit to copy here. You get the mechanics of what worked, the red lines that were crossed (and what they cost), and how to respond when this playbook is used against you.
If you're new to this, start with the fundamentals, from the vote goal to what the research says. Read the fundamentals
Who did it
The operators rarely spoke in detail about the method; what is known comes mostly from reporting, from researchers who monitored public…
The operators rarely spoke in detail about the method; what is known comes mostly from reporting, from researchers who monitored public groups, and from the courts.
Carlos Bolsonaro
Rio de Janeiro city councillor, in charge of his father's social media
According to Jair Bolsonaro himself and the press, he ran his father's digital marketing from 2010 and led the 2018 campaign's social media; in 2026 he was elected senator for Santa Catarina.
SourcesAM4 Brasil Inteligência Digital (Marcos Aurélio Carvalho)
The 2018 campaign’s largest declared vendor
Per Folhapress, it was paid about R$650,000; Carvalho said the firm ran broadcast lists, state groups and groups for reporting fake news, and directed the campaign films in the runoff. At Congress’s fake-news inquiry (CPMI, 2020) he admitted hiring Yacows for a single send to about 8,000 donors announcing a new phone number; he denied mass messaging, and nothing beyond that was proven against the campaign.
SourcesVolunteer group admins
Operators of the network of groups and broadcast lists
Supporters such as José Raimundo, who ran a group of São Paulo voters, and Carlos Nacli, linked to the Nas Ruas movement, who coordinated broadcast lists, described to the press how the network worked.
SourcesPatrícia Campos Mello
Folha de S.Paulo reporter, author of A Máquina do Ódio (The Hate Machine)
On October 18, 2018, she revealed that companies were buying mass-messaging packages targeting the PT (Workers' Party); her 2020 book recounts the investigation and the harassment she faced.
SourcesFabrício Benevenuto (UFMG)
Computer scientist, creator of the WhatsApp Monitor
Built the system that tracked hundreds of public political groups and let journalists and fact-checkers see what was circulating.
SourcesCristina Tardáguila (Agência Lupa) and Pablo Ortellado (USP)
Fact-checker and researcher
With UFMG, they checked the 50 most-shared images in 347 groups and, in The New York Times, asked WhatsApp to restrict forwarding.
SourcesDavid Nemer
Media researcher (University of Virginia)
Followed pro-Bolsonaro groups from March 2018 and described their pyramid structure and three user profiles.
SourcesSuperior Electoral Court (TSE)
Brazil's electoral judiciary; rapporteur Luis Felipe Salomão
On October 28, 2021, it voted 7–0 against annulling the ticket's mandate, but acknowledged the mass messaging and disinformation and set a legal standard for future elections.
SourcesWhatsApp
Platform; Ben Supple, public policy and elections lead
Admitted in 2019 that bulk-messaging companies had violated its terms in the Brazilian election, and limited forwarding to five chats.
Sources
In their own words
Explanations from operators and from those who investigated the case, paraphrased.
Explanations from operators and from those who investigated the case, paraphrased. No message attacking any group is reproduced here.
His son Carlos had handled his marketing and campaigning since 2010 and was the one who started the campaign on social media.
Jair Bolsonaro, Interview with Jovem Pan, cited on Wikipedia (Carlos Bolsonaro article)
Around 2010, searching his father's name on Google, he found only negative results; so he started a blog with family photos to build a positive image.
Carlos Bolsonaro, Account cited on Wikipedia (Carlos Bolsonaro article)
“We directly reached 10,000 people from all over Brazil.”
Carlos Nacli, broadcast-list coordinator, Interview with El País, cited by Folha (DN, Oct 20, 2018)
He was invited to run the group, removed anyone who stirred up trouble and said he kept fake news in check, though he claimed never to have seen any.
José Raimundo, admin of a São Paulo group, Folha de S.Paulo, republished by DN, Oct 20, 2018
6 more quotes
The activity was voluntary but not amateur in its production, and it was coordinated across platforms, with different instructions for YouTube and Facebook.
Mariana Valente (InternetLab), Folha de S.Paulo, republished by DN, Oct 20, 2018
Disinformation flowed down a pyramid: at the top, small groups of influencers created the content; in the middle, the most devoted supporters amplified it; at the bottom, ordinary Brazilians received it all.
David Nemer, Interview with UVA Today, Sep 4, 2019
The influencers were the smallest group, perhaps 5% of users in the political groups, and not the most active in discussion: they produced the content and coordinated protests.
David Nemer, Article in The Guardian, summarized by Exame, Oct 25, 2018
WhatsApp was splitting the country into two big groups.
Cristina Tardáguila (Agência Lupa), Poynter, Oct 17, 2018
Bulk-messaging companies that violated the app's terms of service were active in Brazil's 2018 election.
Ben Supple (WhatsApp), Gabo Festival, Colombia (Olhar Digital, Oct 8, 2019)
There was plenty of evidence that people close to Bolsonaro had worked since 2017 to attack opponents, but there was not enough data on the content and reach of the messages to annul the ticket's mandate.
Luis Felipe Salomão (TSE rapporteur), Opinion at the TSE, Oct 26, 2021 (A Gazeta)
The story
Bolsonaro entered 2018 with little TV time, little declared money and years of social media presence.
Bolsonaro entered 2018 with little TV time, little declared money and years of social media presence. Much of the campaign happened inside closed groups, out of sight of the press and the courts. That has made it hard, to this day, to measure what worked and who paid for what.
- 8 s
- of free TV and radio airtime in the first round
- R$2.5M
- in declared campaign spending, against Haddad's R$37.5M
- 4 of 50
- of the most-shared images in 347 groups were entirely true
- 55.1%
- of valid votes in the runoff, October 28, 2018
From 2018 to the court ruling
- Mar 2018Researcher David Nemer begins following pro-Bolsonaro WhatsApp groups.
- Jul 2018WhatsApp begins testing limits: forwarding to up to 20 chats worldwide and five in India.
- Sep 6, 2018Bolsonaro is stabbed in Juiz de Fora. He does not return to debates and focuses the campaign on social media.
- Oct 7, 2018First round: about 46% of valid votes.
- Oct 17, 2018Lupa, USP and UFMG show that only 4 of the 50 most-shared images in groups were true, and ask WhatsApp for temporary limits.
- Oct 18, 2018Folha reveals that companies were buying mass-messaging packages against the PT, with contracts of up to R$12 million each.
- Oct 28, 2018Runoff: Bolsonaro wins with 55.1%.
- Jan 2019WhatsApp limits forwarding to five chats worldwide.
- Oct–Dec 2019WhatsApp admits bulk-messaging companies violated its terms during the election; the TSE bans mass messaging in Resolution 23.610/2019.
- Oct 28, 2021The TSE votes 7–0 against annulling the ticket's mandate, acknowledges mass messaging and disinformation with no proven link to the campaign, and adopts a legal standard for the future.
Three layers
Research and reporting describe three layers pushing the same content. Only the first was openly run by the campaign.
- Family commandThe accounts of the candidate and his sons, run by Carlos Bolsonaro from home, with videos and livestreams from the candidate.
- Volunteer pyramid networkA few influencers producing, devoted supporters running groups by state and city, and millions of ordinary people receiving and forwarding.
- Paid, opaque mass messagingAgencies that, according to Folha, sold mass sends to businesspeople. The TSE acknowledged the messaging but found no proof that the ticket had hired them.
What isn't known. No study isolates WhatsApp's effect on the 2018 vote.
The groups were encrypted, only public ones could be observed, and the researchers themselves warned that the sample might be skewed. Disinformation also circulated on the other side, though on a smaller scale: of 123 rumors analyzed by three fact-checking agencies over 70 days, 104 hurt Haddad. The stabbing, rejection of the PT and the economic situation also weighed. Testimony from operators is scarce and was given to the press during the campaign.
The terrain
The network fit a very specific environment: a messaging app that worked like a newspaper, rules that did not yet address mass messaging,…
The network fit a very specific environment: a messaging app that worked like a newspaper, rules that did not yet address mass messaging, and a campaign with neither TV time nor money.
Platforms
- WhatsApp groups
- Groups of up to 256 people by state, city or topic, run by volunteers. End-to-end encrypted: only public groups could be monitored by researchers and journalists.
- Broadcast lists
- Only the admin posts, and each list reached up to 256 contacts, who could forward on. This was the channel for seeding videos and images into the network.
- Mass-messaging agencies
- According to Folha (18 Oct 2018), agencies such as Quickmobile, Yacows, Croc Services and SMS Market sold sends of hundreds of millions of messages, at R$0.08–0.40 each, using their own databases segmented by region and income; the report named the retailer Havan among the buyers. Havan’s owner, Luciano Hang, denied it, and in 2020 a first-instance judge ordered Folha and the reporter to pay him damages; the paper appealed. WhatsApp later said bulk-messaging firms violated its terms; the TSE found no proof that people linked to the campaign bought the sends.
- YouTube and Facebook
- Where influencers produced and posted videos, and where groups coordinated actions such as mass subscriptions to the PSL party's channel.
- TV and radio
- Bolsonaro had 8 seconds of free election airtime; Geraldo Alckmin had 5 minutes and 32 seconds, and finished with under 5% of the vote.
Local conditions
- About 6 in 10 Bolsonaro voters read news shared on WhatsApp, against 38% of Fernando Haddad's voters (Datafolha, October 2018).
- 47% of WhatsApp users said they trusted what they received through the app (Datafolha).
- In 2018 a message could be forwarded to up to 20 chats at once; the five-chat limit only went worldwide in January 2019.
- Corporate donations to campaigns were banned; Folha described businesspeople's purchase of mass messaging as an undeclared donation.
- After the stabbing on September 6, Bolsonaro attended no debates and spoke mainly through his own channels.
- A small party (PSL) and a declared campaign budget of about R$2.5 million, against Haddad's R$37.5 million.
The method
This describes what was documented; it is not a recipe.
This describes what was documented; it is not a recipe. The legitimate part (your own presence, volunteers, native formats) is mixed with illegal and harmful practices; the lessons section separates the two.
Three supports
- An owned presence, built early
- Almost eight years of blog posts, profiles and videos before 2018. When TV time was lacking, the audience was already there.
- Distribution by ordinary people
- Content arrived through relatives, colleagues and neighbors, in family and neighborhood groups, where trust runs higher than in the press.
- Emotion and identity
- Voice notes from “people like us”, memes and short videos, designed to travel without context, which also made lying easier.
How a piece of content traveled
Based on Nemer's research, UFMG's monitoring and the 2018 reporting, a piece of content went through roughly these steps.
- Produce Small groups of influencers created videos, images and voice notes, often with no source and no author.
- Seed Broadcast lists and state-level “mother groups” pushed the piece into the network all at once.
- Replicate Admins and devoted members forwarded it to dozens of groups; a single voter was sometimes in 40 groups.
- Filter the bubble Anyone who disagreed was removed. Groups became homogeneous and rarely engaged with the other side.
- Coordinate actions Groups organized tasks on other platforms, such as subscribing to YouTube channels and campaigns against artists' ads.
- Jump outside The piece moved from political groups into family, work and church groups, where it gained the trust of whoever forwarded it.
- Accelerate with money According to Folha, mass sends paid for by companies, using bought databases, multiplied the reach. This is the part the TSE found unlawful.
And again.
What to learn, and what never to copy
The 2018 network showed that real people, on the right channel and in the right format, are worth more than TV time.
The 2018 network showed that real people, on the right channel and in the right format, are worth more than TV time. It also showed the cost of mixing that with hidden money, bought lists and lies.
Never copy
6
- Mass messaging and bought lists
- WhatsApp admitted that bulk-messaging companies violated its terms in 2018. The TSE banned the practice from 2020 and ruled in 2021 that it can lead to annulment of a mandate; Justice Alexandre de Moraes warned that anyone doing it in 2022 would be jailed. The TSE cases also alleged that elderly people’s names and tax IDs (CPF) were used fraudulently to register the SIM cards behind the sends; the court found the evidence insufficient to annul the ticket.
- Off-the-books corporate money
- Folha revealed contracts of up to R$12 million paid by businesspeople, which the law treats as a banned, undeclared corporate donation. The case led to suits at the TSE and an investigation; the court acknowledged the messaging but could not link it to the ticket.
- Disinformation
- Only 4 of the 50 most-shared images in political groups were entirely true. The result was more polarization and, according to Nemer, a radicalization channel that stayed active throughout the government's term.
- Dehumanizing groups
- The groups spread images attacking the rights of LGBT people and women, and a false claim about school materials, according to the BBC and DW. Attacking entire groups harms people and is unacceptable in any campaign.
- Harassing journalists
- Patrícia Campos Mello was targeted with attacks and threats after her reporting. A São Paulo court ordered Bolsonaro to pay her damages over a sexual insinuation, an amount raised in 2022 to R$35,000.
- Opaque attack structures
- The so-called “gabinete do ódio” (hate office), which according to the press was born in the 2018 campaign, became the subject of Supreme Court (STF) inquiry 4.781; in 2022 the Federal Police said the structure had been used by a digital militia.
You can learn
6
- Build your own audience years ahead
- Bolsonaro's social media presence began around 2010. A progressive campaign can do the same ethically: its own channel, a consent-based list and regular posting outside election season. The Janonismo playbook on this site shows the same principle applied by a progressive campaign in 2026.
- Speak in the family-group format
- Short voice notes, vertical videos and images that make sense without context. Use the format, but with true content, sourced and signed by whoever is speaking.
- Give volunteers a clear role
- City and topic group admins were the backbone of the network. Recruit volunteers with consent, give them public rules (no disinformation, no personal attacks) and training to check before forwarding.
- Brief each platform on its own terms
- The network had different instructions for YouTube, Facebook and WhatsApp. Do the same with concrete tasks for actual volunteers, never with secondary accounts or fake engagement.
- Don't depend on TV
- With 8 seconds of airtime, organic distribution made up for the lack of TV. Invest early in your own channels and supporter network; the Janonismo playbook shows how a WhatsApp channel with rules, registered with the electoral authorities, did this in 2026.
- Testimonials from ordinary people
- Voice notes from “people like us” created closeness. Use real testimonials, with the speaker's permission and name, never invented stories.
How to respond
6
- Monitor public groups
- UFMG's WhatsApp Monitor made it possible to see what was circulating in near real time. Follow public groups and trending topics so you know what is coming before it arrives.
- Respond through trusted messengers
- The base of the pyramid distrusted the press. Corrections have to come from relatives, colleagues and local leaders, in the same groups and the same short format.
- Correct without amplifying
- Correcting works and backfire effects are rare, according to the Debunking Handbook 2020; the risk lies in forwarding the false item itself. Lead with the true fact, mention the rumor only once, explain why it misleads and repeat the fact, without resending the original image. Rumors that are not growing can simply be monitored.
- Warn in advance (prebunking)
- Before the final stretch, explain to your network that unsourced images and voice notes will circulate, how to recognize them and where to check. People who have been warned forward less.
- Gather evidence for the courts
- The 2018 suits failed for lack of screenshots, phone numbers and message content. Keep screenshots, originating numbers, dates and links, and report to the electoral court and the platform.
- Demand friction from platforms
- After the five-forward limit, WhatsApp said the volume of forwards fell by 25% worldwide. Push for limits, transparency and researcher access to data.
Practice
Rehearse with the case's missions and situations.
Decide from this side
Six situations for a democratic campaign facing a network like the 2018 one. Choose and see what experience teaches.
Rules
What the law, the courts and the platforms said about what happened in 2018, and the rules that came afterwards.
What the law, the courts and the platforms said about what happened in 2018, and the rules that came afterwards.
- Corporate donations are banned
- Folha described businesspeople's purchase of mass messaging as an undeclared corporate donation, banned under Brazilian electoral law.
- Opt-in lists only
- The law only allows campaigns to use contacts obtained voluntarily. Buying third-party databases is illegal.
- Mass messaging banned
- TSE Resolution 23.610/2019 banned mass messaging without consent from the 2020 elections onwards.
- The TSE standard (2021)
- Mass sending of disinformation through apps can amount to abuse of economic power and misuse of the media and, depending on its seriousness, lead to annulment of a mandate. The court set out five criteria: content, reach, scale, the candidate's involvement and corporate funding.
- WhatsApp's terms
- Automated and bulk messaging violates the terms of service. In 2019 the company was banning more than 2 million accounts a month worldwide and began suing mass-messaging agencies.
- Forwarding limit
- Since January 2019, a message can only be forwarded to five chats at once.
- Harassment has consequences
- In 2022 the São Paulo Court of Justice upheld Bolsonaro's conviction for insulting Patrícia Campos Mello, with R$35,000 in damages.
- Data protection
- Since 2020 the LGPD (Brazil's data protection law) also applies to campaigns in Brazil. In the European Union, the GDPR treats political opinion as sensitive data, and Regulation (EU) 2024/900 allows targeting political ads only with explicit consent and without sensitive data and requires labels and transparency notices on paid ones. Since October 2025, Meta has not accepted political, electoral or social-issue ads in the EU, and Google has stopped serving political ads. See the “Law and platforms in the European Union” section in Fundamentals.
Limits
What this method doesn't do, and when it fails.
- The effect can't be measured
- With encrypted groups and many other factors (the stabbing, rejection of the PT, the economy), no study can say how many votes WhatsApp shifted.
- The courts didn't close the loop
- The TSE acknowledged mass messaging and disinformation but found no proof that people linked to the ticket had paid for it. Part of the evidence never reached the case.
- It wasn't one-sided
- Researchers found misleading content in both left-wing and right-wing groups, although most of the rumors checked hurt Haddad.
- The platform changed the rules
- A five-forward limit, bans on agencies and the TSE prohibition: what worked in 2018 was later shut down or criminalized.
- What came next
- Bolsonaro was declared ineligible by the TSE in June 2023 over events in 2022. On September 11, 2025, the STF sentenced him to 27 years and 3 months for a coup plot, and he was jailed in November 2025. In 2026, Carlos Bolsonaro was elected senator for Santa Catarina and Flávio Bolsonaro lost the runoff to Lula. None of these decisions ruled on the 2018 digital campaign.
Check yourself
Five questions on the mechanism, the red lines and the responses.
Five questions on the mechanism, the red lines and the responses. Each answer comes with an explanation.
Sources
This case study draws on reporting from 2018 to 2026, research on public WhatsApp groups, TSE rulings and statements by the platform.
This case study draws on reporting from 2018 to 2026, research on public WhatsApp groups, TSE rulings and statements by the platform. Pages accessed on October 26, 2026.
31 sources
Who did it
- Diário de Notícias (Folha de S.Paulo): A tática dos eleitores de Bolsonaro no Whatsapp Oct 20, 2018
- Wikipédia: Carlos Bolsonaro
Data
- Poder360: Desde 1994, só Dilma e FHC foram eleitos com maior tempo de TV Sep 4, 2022
- Jovem Pan: Campanha de Haddad gastou 15 vezes mais que a de Bolsonaro Nov 18, 2018
- Wikipedia: 2018 Brazilian presidential election
- Migalhas: TSE rejeita cassação da chapa Bolsonaro-Mourão Oct 28, 2021
- Wikipedia: Trial of Jair Bolsonaro
- Wikipedia: Jair Bolsonaro
Press
- Folha de S.Paulo (via Bem Paraná): Empresários bancam campanha contra o PT pelo WhatsApp Oct 18, 2018
- Correio (Estadão Conteúdo): Bolsonaro dobra aposta em Carlos para campanha nas redes sociais Jan 23, 2022
- Jornal Opção: Livro de jornalista atacada por Bolsonaro analisa a construção da “máquina do ódio” Aug 2, 2020
- DW (via IHU Unisinos): Como 2018 mudou a forma de fazer política no Brasil Oct 30, 2018
- Nieman Lab: WhatsApp limits message forwarding to fight misinformation Jan 2019
- Olhar Digital: WhatsApp admite envio ilegal de mensagens durante as eleições de 2018 Oct 8, 2019
- A Gazeta: TSE tem maioria contra cassação de Bolsonaro e critica WhatsApp nas eleições Oct 28, 2021
- CNN Brasil: TSE rejeita por unanimidade ações que pedem cassação de Bolsonaro e Mourão Feb 9, 2021
- Tecnoblog: Disparos em massa via WhatsApp vão dar cadeia, avisa Alexandre de Moraes Oct 28, 2021
- Brasil de Fato: Justiça mantém condenação de Bolsonaro por ofender Patrícia Campos Mello Jun 29, 2022
- Congresso em Foco: Carol de Toni e Carlos Bolsonaro conquistam as vagas de SC no Senado Oct 4, 2026
- TechCrunch: WhatsApp rolls out new limit on message forwards (25% drop in 2019) Apr 7, 2020
- BNews (Folhapress): Bolsonaro nomeia para transição dono da maior fornecedora da campanha Nov 6, 2018
- Jornal de Brasília: Empresa que atuou em campanha de Bolsonaro diz que ex-funcionário mentiu em CPI Mar 4, 2020
- Poder360: Folha de S.Paulo é condenada a indenizar Luciano Hang, dono da Havan (first instance) Dec 5, 2020
Analysis
- Agência Pública: A Máquina do Ódio, com Patrícia Campos Mello (podcast) Mar 12, 2021
- Poynter: Lupa, USP e UFMG: as 50 imagens mais compartilhadas em grupos de WhatsApp Oct 17, 2018
- BBC News Brasil (via Jornal Grande Bahia): Uma semana dentro de 272 grupos políticos no WhatsApp Oct 19, 2018
- UVA Today: Q&A: What Brazilian presidential election can teach us about fake news (David Nemer) Sep 4, 2019
- Exame: Os 3 perfis de usuários do WhatsApp que ajudam a eleger Bolsonaro (David Nemer, The Guardian) Oct 25, 2018
- Center for an Informed Public (UW): David Nemer on misinformation and WhatsApp in Brazil Mar 4, 2021
- Wikipédia: Gabinete do ódio
- Lewandowsky et al.: The Debunking Handbook 2020 (backfire effects) accessed Oct 26, 2026
Updated 26 October 2026
Next playbook From TV studio to feed