This analysis was performed by Democracy Reporting International with contributions from the FGV’s Law Department.
This report explores gender-based online violent content by analysing public discourse during the political campaign in São Paulo on X and YouTube. Violent discourse targeting women candidates is a recurring issue in Brazilian political campaigns, including in this municipal election. São Paulo stood out among other Brazilian capitals for the amount of political violence that accompanied the campaign. In this context, from the very start of the campaigns, women were the main targets of online violence. In São Paulo’s case, this pattern intensified further in relation to the two major mayoral candidates, Amaral and Helena.
Key findings
- The political campaign for the mayoral election in São Paulo has been intense, marked by significant conflicts between candidates and a concerning degree of gender-based violence targeting the two female candidates, Tabata Amaral and Marina Helena. Out of more than 3, 500 posts containing violent terms collected on X during the first two weeks of campaign, the two were subject to three times as many attacks as their male counterparts. A similar pattern emerged on YouTube. Among the 491 violent comments collected from the platform during a key debate online, although other candidates, especially Pablo Marçal (mentioned in 40.2 percent of these posts), were mentioned more often, women were frequently targeted as well (Amaral in 32.2 per cent of such posts) and subject to the most aggressive attacks, as revealed by qualitative analysis.
- Discussions on X about the candidates ranged from attacks based on political ideology (66 per cent) to explicit misogyny (14.5 per cent). Amaral, the candidate for the Brazilian Socialist Party (BSP), was the primary target of this violent content, facing criticism related to her age, appearance, political affiliations, and background. Other female candidates were also subject to large and disproportionate amounts of such attacks, relative to their overall mentions in the dataset.
- From 1 September, Gazeta and MyNews’s live debate collected more than 4 million views and more than 15, 000 comments. Our analysis of the comments revealed that, along with Amaral, Marina Helena, the candidate of NOVO, and the female journalist moderating the session were subject to gender-based attacks, highlighting how the pervasive nature of misogyny during this political campaign targeted not only female candidates, but also other figures involved in related discussions.
- The final part of this analysis, focused on understanding the nature of these attacks, revealed that the majority of posts and comments seek to undermine or diminish the role of women in politics (90 per cent on YouTube, 80 per cent on X). Other attacks were split between attempts at the dehumanisation, sexualisation, and fetishisation of women, while some were aimed at spreading misinformation about them.
Methodology
This report examines X and YouTube data from the online discourse surrounding the mayoral candidates in São Paulo, the capital of the state by the same name in Brazil. First, we considered any posts on X containing mentions of candidates and violent content (for a total of 3, 566 posts) during the first two weeks of the political campaign (16-31 August). The analysis was limited to this specific time frame due to the Supreme Electoral Court (TSE)’s decision to ban X in the country after the platform refused to adopt measures to restrict disinformation and appoint a legal representative in Brazil. X data was collected using BrandWatch, a consumer intelligence platform that allows users to gather posts using specific queries. Secondly, we considered all comments (15, 548 comments) collected from one of the main live political debates broadcasted on YouTube, in a collaborative video between Gazeta and MyNews channels on 1 September. This data was collected via the YouTube API.
By employing linguistic rules and lists of violent keywords, we filtered the datasets from each platform to retain only the content that exhibited violent characteristics. These lists include terms that can qualify as instances of violence both generally and specifically, depending on the location, context, and the individuals involved. They encompass a range of expressions, from words like “beat up” and “die”, to more local expressions and common derogatory terms used in discussions about women, such as gata, meaning “hot”, “cutie”, vagabunda meaning “homeless”, and fraca, “weak”. In the filtered datasets we performed analyses such as: i) volume analysis, specifically to understand the relative volumes of violent and non-violent content; ii) topic modelling, which is a technique that allowed us to understand the range of the attack arguments; and iii) manual labelling of violent content, to understand the nature of the attacks targeting the two women candidates.
For the last part of the analysis, we classified posts into four categories, as derived from the article Gender and Online Politics: Political Violence and Gendered Disinformation Practices (Curzi et al., 2024), which outlines the major categories of online gender-based political violence, specifically:
- Dehumanisation: A discursive strategy that seeks to “annihilate the existence” of women and/or other marginalised groups in general or in specific contexts, such as in policy decision-making arenas. This strategy can frequently be seen in comments that suggest women don’t belong in politics, or that insinuate they should be removed from the public sphere.
- Undermining/diminishing: According to the study reference above, this is the most common discursive strategy in online gender-based violence, and includes comments containing irony and mockery, with the main goal of disqualifying women and/or other marginalised groups. This strategy commonly involves the use of nicknames and/or puns.
- Sexualisation/fetishism: This refers to comments that are based on the sexualisation and objectification of women. Users reduce women to their physical appearance, and the comments usually contain sexual content and harassment, frequently using terms such as a “whore” or “slut”.
- Gendered disinformation: This refers to the dissemination of false information based on gender stereotypes and norms. It may include content that spreads or reinforces false information about women’s professional and personal lives, with the specific intent to discredit them and their reputations.
Online gender-based violence in the political debate, a concerning issue in Brazil
- Online gender-based violence represents a concerning issue during elections in Brazil, disproportionately affecting women candidates, activists, and voters. This type of violence includes threats, harassment, disinformation campaigns, and the dissemination of false or defamatory content aimed at undermining women’s participation in the political process. A study revealed that 58 per cent of women elected as mayors in 2020 experienced political violence, with many subjected to online threats, slurs, and hate speech . The impact of such violence extends beyond immediate psychological harm, discouraging women from engaging in political activities and undermining democratic processes.
- Recent studies also highlight that women who belong to specific social groups that are culturally and economically marginalised – especially LGBTQIA+, black and indigenous – are usually subject to specific manifestations of violence. This violence is not only targeted, but also intersectional, meaning it is shaped by the overlapping identities and experiences of these women, which combine and deepen the historical violence against these groups. This can be manifested as discrimination, harassment, and physical aggression.
Presenting the political scenario in São Paulo
- As the most populous city in Brazil, São Paulo mayoral candidates’ list for the next municipal election also reflects the city’s diversity of political ideologies. This analysis focuses on the seven main candidates running for mayor, the profiles of whom are briefly summarised below (Table 1).
Table 1: Profiles of Monitored Mayoral Candidates’
Online gender-based violence in the first two weeks of the campaign on X
- X has been at the centre of the debate during the political campaign in Brazil because of the decision of the Supreme Electoral Court (TSE) to ban the platform in the country. This action was taken after X refused to delete accounts that were spreading false political information online and to appoint a legal representative in the country. X recently decided to comply with the orders, however, leading to the platform’s reinstatement several days prior to the writing of this report. Due to X no longer being available in Brazil after the ban, we had to limit the analysis to the period for which data was available, namely, the first two weeks of the municipal campaign (from 16 to 31 August).
- In the first part of the analysis, we performed a volume comparison between all of the posts mentioning one of the monitored candidates. As can be seen below (Graph 1), the candidate with the highest number of mentions on X was Marçal (1, 725, 963 posts), followed by Boulos (956, 408 posts), Helena (285, 882 posts), Amaral (201, 067 posts), Nunes (187, 486 posts), and Datena (56, 038 posts).
Graph 1: Total volume of mentions of candidates on X during the first two weeks of campaign
Source: BrandWatch Elaboration: Democracy Reporting International
- Filtering these mentions for those containing violent content led to a significant reduction of the overall volume of posts (Graph 2). In Graph 3, the data is further broken down to show the mentions with violent content for each candidate.
Graph 2: Total volume of candidates’ mentions with violent content during the first two weeks of campaign on X
Source: BrandWatch Elaboration: Democracy Reporting International
Graph 3: Comparison between mentions with and without violent content by candidates
Source: BrandWatch Elaboration: Democracy Reporting International
- Although male candidates were targeted with the highest volume of violent content in absolute terms, a closer look reveals that the two female mayoral candidates, Amaral and Helena, were more likely to be the targets of violent mentions in relative terms, as can be seen below (Graph 4). Violent content constituted more than 3 per cent of all content mentioning these candidates on X, a substantially higher volume relative to their male counterparts.
Graph 4: Percentage of mentions with violent content received by each candidate compared to mentions without violent content
Source: BrandWatch Elaboration: Democracy Reporting International
- In the next step, we ran a topic modelling analysis to obtain a representation of the content underlying the violent posts identified.
In particular, we identified 4 categories of violent content:
- Ideological attacks: Attacks targeting different points of the ideological spectrum, but predominantly the left, often linking left-wing candidates to communism in a derogatory manner, and using their orientation as a basis for generating political threats. Internal criticism was also present on both the right and the left, focusing on alleged missteps in electoral strategies.
- Misogynistic attacks: Attacks focusing on women candidates’ appearance, undermining their political competence, infantilising them, and questioning their intellectual capacity to be in political positions.
- Attacks on the political elite and political system: Conspiratorial attacks and anti-system sentiment, alleged collusion within the old political establishment, and corruption of the system and political elites.
- Attacks against journalists and traditional media: Attacks on journalists and traditional media, broad criticisms of the female professionals (this was particularly evident in the case of YouTube data, where the journalist who moderated the debate was the target of numerous attacks).
Graph 5: Main topics of violent content on X
Source: BrandWatch Elaboration: Democracy Reporting International
- Graph 5 displays the distribution of these topics, showing that violent posts mentioning candidates on X were mainly in the form of ideological attacks (claiming that candidates are “communists” and/or linking them to property crime and censorship). The second largest category, “Misogynistic attacks, ” highlights the prevalence of gender-based violent content during the political campaign. Amaral has been the main target of such attacks, which have included derogatory remarks about her appearance and accusations of her being weak and fraudulent.
YouTube analysis
- On 1 September, two important digital media outlets, Gazeta and MyNews, collaboratively transmitted a live debate on their YouTube channels (with 979, 000 and 1.06 million subscribers, respectively). Five mayoral candidates were involved: Datena (PSDB), Boulos (PSOL), Marçal (PRTB), Nunes (MDB), and Amaral (PSB).
- Out of 15, 548 comments collected during the live streamed debate on YouTube, 479 contained violent terms or attacks on the candidates (3 per cent).
Graph 6: Comparison between non-violent and violent comments on YouTube
Source: YouTube API Elaboration: Democracy Reporting International
- The topic modelling analysis of violent content in YouTube comments reveals that the main narrative centred around “Attacks on the political elite and political system”, reflecting widespread frustration with and distrust of political institutions. As already seen in the X analysis presented above, this narrative was closely followed by “Misogynistic attacks”, highlighting the persistent hostile communications against female candidates.
- It is important to note that, even within the topic “Attacks against journalists and traditional media”, this content is predominantly aimed at female journalists, often invoking common gender stereotypes.
Graph 7: Main topics in violent content discussed on YouTube comment section
Source: YouTube API Elaboration: Democracy Reporting International
- Posts containing violent words often made specific mention of one or more candidates, both present and absent from the live debate. When analysing the mentions per candidate, we see that Marçal garnered the most attention (40.2 per cent). Interestingly, most of these mentions framed him in a positive light.
- In contrast, while Amaral ranked second in terms of overall mentions (32.2 per cent), she was subjected to numerous direct and aggressive attacks, highlighting a stark contrast in the sentiment of discourse surrounding the most frequently discussed male and female candidates.
Graph 8: Number of violent comments on YouTube mentioning one or multiple candidates
Source: YouTube API Elaboration: Democracy Reporting International
Understanding the nature of the attacks
- To gain a deeper insight into the nature of the attacks directed at the two female mayoral candidates, we conducted manual labelling of the X posts and YouTube comments explicitly mentioning them. This labelling was based on the classification outlined in the methodology section.
- Among the four classifications we used, the category of “Undermining/diminishing” was by far the most frequent. Over 90 per cent of YouTube comments derided female candidates using nicknames and/or puns, particularly in the case of Amaral. On X, posts falling into this category accounted for 80 per cent of the total.
Graph 9: Understanding the nature of attacks aimed at Tabata Amaral and Marina Helena
Source: YouTube API Elaboration: Democracy Reporting International
- Attacks against Amaral included:
-
- Gendered disinformation:
-
-
- Here, we identified comments whose main characteristic was spreading or reinforcing false information about Tabata’s professional and personal life, in order to discredit her, raising questions about her character and reputation.
- For instance, she is often accused of being associated with an important businessman, Jorge Paulo Lemann, due to the fact that she received financial support from him to complete her degree in Astrophysics and Political Science at Harvard University. Frequently delivered in a conspiratorial tone, this information has been strategically evoked to undermine her autonomy and agency, by suggesting that Tabata is only a “billionaire’s puppet”, and that she represents the alleged hidden interests of an elite. In other words, these comments insinuate that, as a woman in politics, she must be controlled by a more experienced and powerful man, whose interests she will supposedly defend if elected.
- Users also commented that Tabata allegedly supported the censorship of the social media postings of Marçal, who is another very controversial mayoral candidate. In this context, Tabata is considered by some users an “enemy of freedom of expression”.
-
-
- Undermining/diminishing:
- There were also many comments calling Tabata “spoiled” and a “preppy girl” (in Brazilian Portuguese, “patricinha”), due to her age (she is 30 years old). Based on gender stereotypes, these comments aim to suggest that Tabata is weak, fragile, and unprepared for the assignment.
- With the intention to undermine her, users use a pun to call her by nicknames such as “Batata” or “Tapada” (in Brazilian Portuguese, “batata” means potato, while “tapada” means stupid or idiot).
- Undermining/diminishing:
-
- Sexualisation/fetichism:
- Tabata’s age is frequently used to justify comments that sexualise her as a young attractive woman, while denigrating her abilities and even her character. Users often refer to her as “hot” and “cute” (“gata”, “gostosa”, “bonita”), despite, they claim, her “lack of intelligence” and her “being boring”.
- Sexualisation/fetichism:
-
- Dehumanisation:
- In this category, user comments are very aggressive, swearing and using foul language. The content includes threats, wishes for harm or disgrace, and frequent claims that the individuals targeted deserve such treatment. The most common insults involve sexualised language (including crude terms such as “whore” or “slut, ” and social degradation, e.g., “homeless”). These terms are commonly used in isolation, without context or further explanation, reflecting a pattern of quick-hit insults.
- Dehumanisation:
- Attacks against Helena:
-
- Although she was mentioned less frequently than Amaral, Helena was subject to comments characterising her as “weak”, and suggesting that she would not garner many votes.
-
- We also identified a small number of comments with sexual content related to Helena.
- Graph 10 shows the most frequent words used in the posts and comments aggressively targeting Amaral or Helena. We can see that the majority of the attacks (in yellow) are directed at Amaral, using nicknames or puns (Tabata – batata) when not directly insulting her (fraca, “weak”; ridícula, “ridiculous”; burra, “dumb”).
Graph 10: Most frequent words in posts and comments targeting Tabata Amaral and Marina Helena
Source: Brandwatch and YouTube Elaboration: Democracy Reporting International
About the Media and Democracy Project
The School of Communication, Media, and Information of the Getúlio Vargas Foundation (FGV ECMI) and the European Union (EU) have established a partnership to collaborate on strengthening Brazilian democracy and the integrity of digital spaces. The initiative, known as the Media and Democracy Project, will be conducted in collaboration with the fact-checking platform Lupa, “Democracy Reporting International” (DRI), the German centre for public debate analysis. The project focuses on debates concerning gender issues, racial inequality, minority rights, hate speech, and other important issues in Brazilian politics. The project encompasses monitoring and analysis of digital media, fact-checking, and the establishment of a council as a public space for promoting positive dialogue, involving researchers, civil society, governmental representatives, and platform stakeholders.
