Digital Democracy

Digitalization and the Public Sphere in Brazil


Scientific Claims about Covid-19 on Twitter

Download (PDF)



The study analyzes publications on Twitter that claim scientific status in order to bring forth arguments of authority in messages regarding Covid-19, including the adhesion or criticism of protective measures, the application or non-application of vaccines, degree of danger of the virus, among others. The 3.3 million posts that compose the corpus of the research were published between January and May 2021. Of the four identified clusters, the one that garnered the most engagement mainly dealt with the defense of early treatment and was positioned in the right-wing conservative field in alignment with the government of Jair Bolsonaro. The other three clusters reacted to this, including health professionals, sanitary authorities, epidemiologists, journalists, as well as progressive and left-wing influencers. In the group aligned with the government, the longevity of links that resort to science in order to argue about Covid-19 was 150% higher than in the other three clusters identified, which indicates that, despite not always grounding themselves in scientific parameters in order to base their arguments, the right-wing conservative cluster explores the status of science in order to defend their point of view.

Covid-19; Twitter; science; pseudoscience; public sphere.


  • The cluster that attracted the highest volume of interactions (41.5%) on Twitter was composed of users aligned with the conservative right-wing field, defending early treatment protocols for Covid-19. This is not the first, but the third cluster in terms of quantity of profiles connected with each other. Therefore, even with 21.5% of the profiles, this group worked separately as a vector of messages that preach the efficacy of early treatments. The most populous cluster, with 29.6% of profiles, was composed of health professionals, scientists and sanitary authorities that allude to the lack of evidence of the early treatment protocols, among other topics, and is the third in terms of interactions (11.6%). The second cluster in interactions (34.5%) and profiles (24.9%) is represented by the left-wing field, with criticisms directed to the federal government. In turn, the smallest cluster (9.5% of profiles and 7.7% of interactions) is composed of epidemiologists, journalists and associates of infectiology who also criticized the management of the pandemic by the Jair Bolsonaro administration.
  • Among the five domains with the most mentions, responses and retweets by the cluster that defends the early treatment, all are vehicles of the conservative right-wing media and hyperpartisan. Domains of websites of the traditional press occupied the five first positions of the cluster connected to the health professionals and sanitary authorities, sharing the space with websites editorially oriented to the left, in groups of opposition to the government, and to the right, in the group more connected to the center-right wing.
  • The average lifespan of links in circulation about the subject was of 100 hours in three of the four clusters. An exception happened in the cluster supporting early treatment, where the duration increased to 250 hours. Considering the ten most lasting links, the cluster defending early treatment is the only one where controversial URLs that propagate pseudoscience could be identified. Those are anonymous websites that are constantly refuted and invalidated by the scientific community. The most durable link circulated for 159 days and it is an example of this practice in defense of the early treatment, as well as others with longer lifespan. In turn, the three less durable links circulated for around 85 days and are all from the media outlets.
  • The cluster favorable to the early treatment was the one that isolated most exclusive/predominant domains. That means that, in 44% of the links, 90% of the mentions, responses and retweets occurred inside the cluster itself. A significative percentage of isolation of informative sources occurred in the cluster that gathers users who criticize the government and who are aligned with the left wing, where 31% of the links circulate predominantly.
  • It is also worth noting, also, that the cluster that defends the efficacy of the early treatment was the group that most shared six of the fourteen scientific domains that were shared among all of the clusters.


1) Analysis of the Map of Interactions

In the general debate about Covid-19, more than 3.3 million of posts on Twitter that claim some scientific foundation for their respective assertions and stances were identified.

Map of Interactions in the Debate over Covid-19 on Twitter, with Scientific Claims Analysis
period: from January 1st to May 31st, 2021

Source: Twitter | Elaborated by: FGV DAPP

. Source: Twitter | Elaborated by: FGV DAPP

Orange ‒ 29.6% of Profiles | 11.6% of Interactions
Mobilized by health professionals and other sanitary authorities, the group contests the recommendations of the supposed early treatment, alluding to the lack of scientific evidence on the efficacy of the respective medicine. Furthermore, posts raise doubt to the reports of success of the medicine in the treatment of the disease and reinforce alert that Covid-19 does not behave as a normal flu, insisting in other demands, such as the use of masks and hand sanitizers.

Light Blue ‒ 24.9% of Profiles | 34.5% of Interactions
Group lead by left-wing politicians, celebrities and social activists that oppose President Jair Bolsonaro attributed the delay of the immunization of the population against Covid-19 to the lack of trust of the federal government in the vaccines and its inertia in the acquisition of them. At the same time, while insisting in the urgency of the vaccination, posts criticize the favoring, by the government, to medicine that compose the labeled “early treatment” whose efficacy against the disease is not scientifically backed.

Lilac ‒ 21.5% of Profiles | 41.5% of Interactions
Orbiting right-wing politicians, conservative bloggers and digital influencers, the group defends protocols of early treatment against Covid-19. Posts resort to personal reports of recuperation from the disease to studies and declarations from international specialists, as well as their approval in foreign countries, in the effort of basing the efficacy of the medicine. Some posts also question the accusation of this posture as being negationist, with the argument that negationism would be the impediment of access of the population to this early treatment.

Green ‒ 9.5% of Profiles | 7.7% of Interactions
Mobilized by epidemiologists, journalists and infectology associations, this group comments on the inefficacy of the use of ivermectin and chloroquine as early treatment against Covid-19. The group also criticizes the federal government for insisting in the production and propaganda of these medicine, as well as depreciating the investment in vaccines and in a plan of vaccination for Brazilians. Posts label the government’s posture as negationist and highlight cases of supporters of the government who are in favor of the early treatment, but who vaccinated in other countries.

Domains with most Mentions, Responses and Retweets by Cluster Analysis
period: from January 1st to May 31st, 2021

Source: Twitter | Elaborated by: FGV DAPP

. Source: Twitter | Elaborated by: FGV DAPP

The graph above presents the domain with most mentions, responses and retweets by cluster. In the Lilac cluster, the presence of hyperpartisan and conservative websites is predominant — such as and They tend to demand minimum participation of the State and, throughout the pandemic, published articles in support of the federal government. Another group that draws attention by the presence of partisan websites is the Light Blue group, which brings forth news channels from the left wing, whose identification is marked by progressive ideals and defense of the democracy — highlights are and This group presents criticisms to the federal government in regards to the adopted measures to fight the pandemic and in the insistency on the narrative in favor of drugs of the early treatment.

The groups Green and Orange are domains marked by traditional websites of communication — for example:,, — which spread information on vaccines, contest the use of preventive medicine against Covid-19, and emphasize the high investment of the Brazilian government in the purchase of chloroquine. It is worth noting that the link — with wide circulation in the Green domain — despite being an informative channel with public editorial positioning to the right wing, spread information emphasizing the inefficacy of the drugs used as preventive treatment against Covid-19 and the possible damage that such medications can bring to the health.

Seeking more detailing of the clusters, the graph below presents the links that obtained more mentions, responses and retweets in each one of them. Among the domains analyzed, the Lilac cluster is presented as the most expressive, with highlights to the website

2) Analysis of Lifespan of Links

This section examines the lifespan of circulation of the links used for the scientific claims on Covid-19 on Twitter. For that, we took into consideration the difference, in hours, between the last and the first time the link appeared in a tweet in the analyzed period, independently of how many times it was later shared by some of the users. For that, we used a group of 40 links with the ten longest lifespans in each cluster with at least two shares. Next, we examined the accumulation of days of circulation of the links. In this way, it was possible to advance the understanding of the habits of sharing of content anchored in links, that evidence a typical headline text (clickbait or not), according to the genesis of functioning of different clusters. In the two observed levels, the data composed of links belonging to the Lilac cluster differ from the rest.

This analysis evidence that the average of duration of the links is around 250 hours in the Lilac cluster, 100 hours in the Light Blue and Green clusters, and 89 hours in the Orange one. The Lilac cluster, whose average is more than double, is the one that attracted most interactions and defended the early treatment in the online conversation on Twitter. The other three clusters point to an average pattern of around 100 hours, with average time being a little shorter in the largest grouping (Orange) who gather users connected to the field of Health who refute the idea of early treatment. The calculation of the median, in turn, signals the central tendency of temporal distribution, which we also applied in this analysis in order to control distortions due to the accentuated volume of data. Thus, in all of the clusters, the central lifespan of the links was lower than 10 hours, with higher duration in the Lilac grouping and lesser duration in the Orange one.

Lifespan (in hours) of Tweets with Scientific Claims on Covid-19 Analysis
period: January 1st to May 30th, 2021

Source: Twitter | Elaborated by: FGV DAPP

. Source: Twitter | Elaborated by: FGV DAPP

Therefore, the most lasting URL circulated for 159 days on Twitter, but all of the ones that are reported in the sample of this research were shared for at least 8 days (Figure 4). Among the ten URLs with most days of circulation, five are from the Lilac cluster, while 4 are from the Light Blue and one is from the Green clusters, which evidences the existence of endogenous ecosystems where circulate information that is controversial and refuted by the scientific mainstream and national and international health authorities.
In order to illustrate the strength of this topic in the Lilac online community, the URL that is on the top of the list is, with 159 days of lifespan. It is a website that systematizes and publicizes studies in defense of the early treatment against Covid-19. The account on Twitter was banned and the project accuses the platform of censorship. The scientific results in favor of early treatment disclosed by this page, which were used as source of information by other channels classified as negationist and hyperpartisan, have already been checked and disproven by the major fact-checking initiatives in Brazil.

The project Aos Fatos has been alerting that this anonymous website discloses meta-analysis, which are synthesis of results of academic researches, but with crass errors of methodology and with no scientific severity, since they group “incomparable studies and applied the drug in different doses and control groups”. The website was mentioned, alongside another connected to it, in the “Manifest for Life”, an advertisement signed by doctors who are in favor of the early treatment and publicized in newspapers of great circulation in Brazil in February of 2021. The fact-checking agency Lupa considered that the text of the manifesto is based on false data, in inconclusive studies and in analysis without scientific approval., which appears as the third most lasting link in the Lilac cluster, with 153 days, is one of the most interconnected URLs in the website, which is a constant source of attention from scientific institutes who are concerned by the widespread of false data and information in strategic networks.

nother URL of primary source,, keeps its prolonged circulation in online communities in favor of the early treatment in Brazil and in the world. With 146 days of circulation, the way that the study of 2005 has been used in posts of social media was rebutted by the fact-checking agency Reuters. By verifying the case, Reuters informs that the “study that proves efficacy of the chloroquine for Covid-19” has a verdict the status of false. The verification clarifies that the experiment was performed in animals and not in humans, having the focus on Sars-CoV (Severe Acute Respiratory Syndrome), a different disease from Sars-CoV-2, and it was a source of concern in that time. The agency reiterates that the studies are inconclusive on the efficacy of the hydroxychloroquine and chloroquine drugs, and that the adverse effects have been alerted. Additionally, the URLs of the homepage, a website owned by a conservative and evangelic media group known for publicizing deceitful content, and the blog , are on the list of the ones who circulate the most in the Lilac cluster, both with 149 days of duration.

3) Analysis of Predominance of Domains

This section investigates the pattern of sharing of informative sources of the clusters identified in the map of interactions. For that, we cross examined the domains with the sharing of the clusters. Among the publications that made scientific claims in the debate on Covid-19, we identified 1156 domains that generated around 300 thousand mentions, responses and retweets.

The first analysis realized uses a metric of exclusivity/predominance of the domains shared by profiles of each cluster. We use, alongside the idea of exclusivity, the proposal of predominance, because the clusters can share domains with whom they disagree, in posts that criticize them. Therefore, a domain was defined as exclusive/predominant when it obtained 90% of its shares in just one cluster.

Exclusive/Predominant Domains by Cluster Analysis
period: January 1st to May 31st, 2021

Source: Twitter | Elaborated by: FGV DAPP

. Source: Twitter | Elaborated by: FGV DAPP

We observe that the Lilac and Light Blue clusters present the most variety of distinct domains, with little more than 740 unique domains. These clusters are also the ones who presented the largest number of exclusive/predominant domains, with 44% and 31%, respectively. The Green and Orange clusters, besides presenting a lower number of shared domains (461 and 291, respectively) also obtained a very low rate of exclusive/predominant domains.

This data helps to better understand the dynamics established by the clusters identified in the analysis. The Lilac group is presented as the most isolated in relation to the informative sources, which converges with the observation realized in the previous section, which refers to the strong presence of hyperpartisan websites among its links with most mentions, responses and retweets. The presence of vehicles of the traditional press, which is increasingly intensified throughout the groups Light Blue and Green to Orange, seems to relate to a more integrated pattern in relation to the more usual sources of information in the political discussion.
Still exploring the relation between exclusivity/predominance and the integration of domains shared by the clusters, we analyzed the domains that were shared between all of the clusters. Of the 1156 identified domains, only 156 were shared at least once among all of the clusters, and, from these, more than 50% of the domains re from vehicles from the traditional press.

In this list of shared domains among all of the clusters, 14 domains from scientific magazines and international sanitary authorities are found. Figure 6 shows the distribution of sharing of these domains by cluster.


The study continues the efforts to investigate how the public debated around discourses that mobilize the scientific statute mediated by platforms of social media in the context of the Covid-19 pandemic happened in Brazil. From January 1st to May 31st of 2021, we analyzed a corpus that gathered more than 3 million posts on Twitter, encompassing more than 42 thousand links, the focus of analysis in the study. Generally, it was possible to observe that the scientific claims had been strongly directed towards the topics of early treatment against Covid-19, more specifically, the debate was marked by stances that were either favorable or against to the use of drugs of the early treatment, with the sanitary measures and vaccination being topics that were tackled peripherally. The data confirm, yet, the predominance of profiles and groups aligned with the federal government, in presence and engagement. This segment recurs to arguments allegedly scientific in order to validate and defend treatments and medications repeatedly refuted by the scientific community. The expressive presence of channels of the hyperpartisan media — frequently pointed out as responsible by the circulation of deceitful information — reinforces the emphasis in the political dispute around the topic. Furthermore, the links originated in scientific magazines, when mobilized, barely follow the nexus that reigns the scientific field, or are originated in pseudoscientific websites, being constantly proven false by the press and stablished scientific institutes. In the other observed groups, the dispute around the topics also remains predominantly limited by sources that take political-partisan stances, such as when informative channels of the left or right-wing fields, which have loyal audiences, mobilize a criticism to the federal government or to the conspiracy theories. Also, it is highlighted the isolation of informative sources mobilized in the distinct groups, indicating that the information circulates very restrictively in specific groups — this endogenous characteristic is specially noted in the community that defends the efficacy of the early treatment, but also in the left-wing cluster. Thus, it is possible to understand that the claims of scientific legitimacy were incorporated in the discursive dispute that characterized the political scenario of the country, but it manifests itself particularly around a dissonant cluster that denies the scientific knowledge established to defend the cause of the early treatment.


ARAUJO, R. F.; DE OLIVEIRA, T. M. Desinformação e mensagens sobre a hidroxicloroquina no Twitter: da pressão política à disputa científica. AtoZ, n. 9, v. 2, p. 196-205, 2020. Available at: Accessed on: 19 jul. 2021.

BOURDIEU, P. Usos sociais da ciência. São Paulo: Unesp, 2003.

BORDIEU, P. Para uma sociologia da ciência. Lisboa: Edições 70, 2004.

CHALMERS, A. F.; FIKER, R. O que é ciência afinal? São Paulo: Brasiliense, 1993.

JENKINS, H. Cultura da Convergência. São Paulo: Aleph, 2013. E-book.

KUHN, T. The structure of scientific revolutions. Chicago: The University of Chicago Press, 1970. Available at: Accessed on: 19 jul. 2021.

LEWANDOWSKY, S.; GIGNAC, G. E.; OBERAUER, K. The role of conspiracist ideation and worldviews in predicting rejection of science. PLoS ONE, v. 8, n. 10, p. 1-11, 2013. Doi:

MACHADO, C. C. V. et al. Ciência contaminada: analisando o contágio de desinformação sobre coronavírus via YouTube. [S.l.]: Laut, INCTDD, Cepedisa, 2020a. Available at: <>. Accessed on: 19 jul. 2021.

MACHADO, C. C. V. et al. Scientific self isolation: international trends in misinformation and the departure from the scientific debate. [S.l.]: Laut, INCTDD, DFRLab, Vero, 2020b. Available at: <>. Accessed on: 19 jul. 2021.

OLIVEIRA, T. Desinformação científica em tempos de crise epistêmica: circulação de teorias da conspiração nas plataformas de mídias sociais. Fronteiras – estudos midiáticos, n. 22, v. 1, p. 21-35, 2020. Doi:

RECUERO, R. Introdução à análise de redes sociais online. Salvador: EDUFBA, 2017.

RUEDIGER, M. (coord.). Nem tão #simples assim: o desafio de monitorar políticas públicas nas redes sociais. Rio de Janeiro: FGV DAPP, 2017. Available at: Accessed on: 19 jul. 2021.


Coordination of Research
Marco Aurelio Ruediger
Amaro Grassi

Victor Piaia
Sabrina Almeida
Tatiana Dourado
Danilo Carvalho
Marcela Canavarro
Dalby Dienstbach Hubert
Maria Sirleidy Cordeiro

Technical Review
Renata Tomaz

Graphic Project
Luis Gomes
Daniel Cunha


Sign up for our newsletter