• The debate on Twitter between November 1st and 8th had 124.9 thousand mentions about the elections in Rio, with 76 automated profiles which were responsible for 1,918 retweets. Eduardo Paes was the main target of negative messages, mentioned in 44 automated publications and the target of 30% of the attacks (organic and bots) on the platform in the general debate;
• The largest group using automation is the left-wing group, aligned with the campaigns of Benedita da Silva and Renata Souza with 873 suspected publications. However, the group focused more on positive content about Benedita;
• Martha Rocha’s support group also used automated accounts to propagate messages (positive and negative). That group was responsible for 356 posts, including criticism towards Paes and Crivella, but mostly replicating posts made by official accounts and praising Martha – who was also the target of attacks made by other groups, representing 8% of the political attacks in the discussion;

Since 2014, specific disinformation structures have regularly appeared on Twitter as elections approach, promoting candidacies and, especially, attacking rivals. This time, although with a lower volume than that observed during the 2018 electoral period, the race in Rio de Janeiro again shows signs of automated profiles (bots) acting on the platform. A monitoring performed by the Department of Public Policy Analysis of Fundação Getulio Vargas detected 76 automated accounts between November 1st and 8th, which made 1,918 retweets inside a general database of 124.9 thousand posts. The current leader of the voting intention polls, Eduardo Paes, is the main target of bot attacks, followed by Martha Rocha, who is running in the second round against Marcelo Crivella.

Over the past years, automated disinformation activity has adapted to detection processes and become more sophisticated, making it difficult to measure the number of accounts showing non-human behavior such as fast reproduction of messages and use of non-validated automated publication software. This time, in Rio, the main ones responsible for the activities of suspected profiles are group defending left-wing candidacies – this does not necessarily imply that the candidates or campaigns themselves are responsible for the activities of these profiles. The debate group favoring the campaign of Benedita da Silva and Renata Souza, for instance, had 712 bot interactions, and the group supporting Martha Rocha had 356. The right-wing group, especially those promoting Crivella’s campaign, had a lower level of bot activity with only 39 automated interactions.

Eduardo Paes was directly mentioned in a negative way in 44 publications replicated by bots (especially coming from the center left group), while Martha was mentioned more often in automated posts (174 times), although mostly to promote her candidacy. Paes was the target of 30% of the negative publications on the network coming from automation and from organic repercussion of attacks – especially from the group aligned with the federal government and Crivella –, much higher than the second place, Martha Rocha, who was targeted by 8% of the critical publications, mostly by left-wing groups. Benedita appeared in 873 publications made by profiles whose activity is similar to bots, with the use of accounts in the left-wing group to retweet posts from her official account, defend the candidate, and replicate content such as election polls.

Automated attacks and negative messages in the Rio de Janeiro elections
Analysis date: November 1st to 8th

The map of interactions below shows the location of profiles showing suspicious behavior on Twitter as black dots. There is a small number of accounts currently operating, and the blue group, supporting Crivella (and, to a lesser extent, Luiz Lima) has the most profiles, but produced a low volume of interactions. The pro-Benedita group also has a small number accounts, but they produced a high number of posts. Other profiles also appeared sparingly in the discussion, without participating in any of the major debate networks. These profiles mostly emphasized Paes.

Map of interactions with automated profiles in the Rio de Janeiro elections
Analysis date: November 1st to 8th

So far, the low participation of bots in the municipal debate about Rio de Janeiro is in sharp contrast to the first round of the previous elections, in which automation was a major tool of political activities from the beginning, This year, due to the higher fragmentation of campaigns on different digital platforms, with a predominance of WhatsApp and Youtube, the strategies accompanied the changes in the online context. Most attacks were organized based on profiles that do not directly indicate automation (at least in terms of their behavior), but continue to have an impact and are relevant in the execution of virtual attacks. However, an increase in the number of bots is to be expected in this final stretch of the first round in order to build up the capillarity of political rejection on Twitter. It is possible that these profiles are saving a wider action for the second round, when the decision process will be effectively determined, especially in cities where with a high level of polarization in opposite fields, foreshadowing the 2022 elections.

In São Paulo, automated accounts focused on criticizing Covas and defending Russomanno

In São Paulo, the context is very similar to that of Rio de Janeiro. With a larger number of identified accounts, 133 automated profiles were responsible for 2217 retweets. Similarly, coordinated actions were performed by profiles that do not directly indicate automation, but still have an impact.

Map of interactions with automated profiles in the São Paulo elections
Analysis date: November 1st to 8th

The blue group had the highest rate of bot participation, and its main target was PSDB candidate Bruno Covas. The posts sought to negatively associate the current mayor’s candidacy with Governor João Dória. This same group also tried to promote Russomanno’s candidacy, praising the candidate in their messages.

On the other hand, the automated accounts identified in the red group focused on amplifying the existing organic reach, with publications supporting the candidate Guilherme Boulos, with no partisan association or specific targets, in addition to teasing the candidate Arthur do Val, known on the social networks as “Mamãe Falei”.