On the Globalization of the QAnon Conspiracy Theory Through Telegram Mohamad Hoseini,1,* Philipe Melo, 2, * Fabricio Benevenuto, 2 Anja Feldmann, 1 Savvas Zannettou 1 1 Max Planck Institute for Informatics 2 Federal University of Minas Gerais fmhoseini, anja, [email protected], fphilmelo, [email protected] Abstract DC in 2016 (Siddiqui and Svrluga 2016). Taken all together, there is a pressing need to understand how these conspiracy QAnon is a far-right conspiracy theory that became popu- theories spread online and how users are radicalized from lar and mainstream over the past few years. Worryingly, the the exposure to conspiratorial content. QAnon conspiracy theory has implications in the real world, with supporters of the theory participating in real-world vio- One conspiracy theory that is considered very persuasive lent acts like the US capitol attack in 2021. At the same time, and attracted high engagement from people is the QAnon the QAnon theory started evolving into a global phenomenon conspiracy theory. The theory alleges that a secret cabal of by attracting followers across the globe and, in particular, in people run a child sex-trafficking ring and is working against Europe. Therefore, it is imperative to understand how the Donald Trump (as the US president by the time the theory QAnon theory became a worldwide phenomenon and how emerged). Over the years, the conspiracy theory attracted this dissemination has been happening in the online space. many new followers across the globe and essentially evolved This paper performs a large-scale data analysis of QAnon into a cult. Worryingly, the followers of the QAnon conspir- through Telegram by collecting 4.5M messages posted in 161 acy theory started making threats or participate in violent QAnon groups/channels. Using Google’s Perspective API, real-world incidents (e.g., Capitol attack in 2021 (Gilbert we analyze the toxicity of QAnon content across languages 2021)), hence highlighting the impact that the conspiracy and over time. Also, using a BERT-based topic modeling theory has in the real world (Beckett 2020). approach, we analyze the QAnon discourse across multiple languages. Among other things, we find that the German Motivated by the negative impact that QAnon has in the language is prevalent in QAnon groups/channels on Tele- real world, mainstream platforms like Facebook, Twitter, gram, even overshadowing English after 2020. Also, we find and YouTube, started moderating and removing QAnon- that content posted in German and Portuguese tends to be related content. Due to this, QAnon supporters sought new more toxic compared to English. Our topic modeling indi- online “homes” in less-moderated platforms and migrated cates that QAnon supporters discuss various topics of inter- to other platforms like Parler and Telegram. At the same est within far-right movements, including world politics, con- time, the QAnon conspiracy theory became a global phe- spiracy theories, COVID-19, and the anti-vaccination move- nomenon; the QAnon conspiracy theory started accumulat- ment. Taken all together, we perform the first multilingual ing new followers across the globe, particularly in European study on QAnon through Telegram and paint a nuanced countries like Germany and Spain (Scott 2020). Overall, it is overview of the globalization of the QAnon theory. imperative to understand how QAnon evolved and became a global phenomenon over time. To do this, we use Telegram Introduction as the source of our study for two reasons. First, anecdotal evidence suggests that QAnon followers migrated to Tele- The spread of conspiracy theories online is not a new prob- gram after bans on other platforms (Dickson 2021). Second, arXiv:2105.13020v1 [cs.CY] 27 May 2021 lem. For instance, conspiracy theories related to the 9/11 at- Telegram is a rapidly growing platform with worldwide cov- tack (e.g., the attack was, in fact, a controlled demolition) erage, hence it is the ideal platform for effectively studying and the Sandy Hook shooting (e.g., the shooting was staged) the QAnon conspiracy theory across the globe. were extensively disseminated on the Web. Over the past few years, however, we witness an explosion in the spread and Research Questions. We focus on answering the following popularity of conspiracy theories on the Web. More worry- research questions: ing is that newer conspiracy theories seem to have a more • RQ1: How does the QAnon community evolve on Tele- negative impact on the online and offline world. People get gram over time and across languages? How toxic is the radicalized online from the continuous exposure to conspir- QAnon community on Telegram? atorial content and then perpetrate violent acts in the real • RQ2: How popular is QAnon content on Telegram? world. For instance, the Pizzagate conspiracy theory was • RQ3: What are the main topics of discussion on QAnon- the driving factor for a shooting at a pizzeria in Washington related groups/channels on Telegram? Are there differ- ences across languages? *Authors contributed equally To answer the above-mentioned research questions, we perform a large-scale data collection and analysis of QAnon- was an American government official with classified infor- related groups/channels on Telegram. Overall, we collect mation about plots against then-President Donald Trump. 4.5M messages shared in 161 Telegram groups/channels be- Subsequently, “Q” continued disseminating cryptic mes- tween September 2017 and March 2021. Using Google’s sages about the QAnon conspiracy theory (called “Q drops”) Perspective API, we investigate the toxicity of QAnon con- mainly on 8chan. The QAnon conspiracy theory accu- tent on Telegram and assess whether the movement is be- mulated many followers on fringe Web communities like coming more toxic over time and whether there are substan- 4chan/8chan and mainstream ones like Facebook (Sen and tial differences across languages. Also, using a multilingual Zadrozny 2020) and Twitter, especially after then-president BERT-based topic modeling approach, we study the QAnon Donald Trump retweeted QAnon-related content (Nguyen discourse across multiple countries/languages. 2020). QAnon followers usually use their motto “Where We Main findings. Our study provides some key findings: Go One, We Go All” (or simply wwg1wga) to tag content • We find that the QAnon movement increased substantially related to the QAnon conspiracy theory on the Web. during 2020 and 2021 on Telegram, which is likely due Over the past few years, followers of the QAnon conspir- to people migrating from other mainstream platforms that acy theory made violent threats or were linked with several take moderation actions against QAnon content. Further- incidents of real-world violence (Beckett 2020), with the more, by comparing content across languages, we find Federal Bureau of Investigation (FBI) labeling it as a po- that German content overshadowed English in popularity tential domestic terrorist threat (The Philadelphia Inquirer during 2020 and 2021 (RQ1). 2020). The last straw of these incidents happened on January • By analyzing the toxicity of QAnon-related messages, we 6th, 2021, when supporters of the QAnon conspiracy theory find that toxicity is rising over time (almost 2x times more attacked the US capitol in an attempt to overturn Donald toxic messages in March 2021 compared to September Trump’s defeat in the 2020 US elections by disrupting the 2019), while at the same time, we find substantial dif- Congress that was in the process of formalizing Joe Biden’s ferences across languages, with German and Portuguese victory (Gilbert 2021). Due to these threats and violent in- being the most toxic languages. (RQ1). cidents, mainstream platforms like Facebook (BBC 2020), • Our analysis indicates that QAnon content on Telegram Twitter (BBC 2021), Reddit (Zadrozny and Collins 2018), is popular and can be viewed by many users. Specifically, and YouTube (Zadrozny and Collins 2020) started monitor- we find one order of magnitude more views and forwards ing content related to QAnon and actively removing groups, on QAnon content than a baseline dataset containing a subreddits, and users that are related to the QAnon conspir- set of political oriented groups. Also, we find that over acy theory. Naturally, following these content moderation time QAnon content is becoming even more popular and interventions, supporters of the QAnon conspiracy theory reaches an increasing number of Telegram users (RQ2). flocked to other fringe Web communities, with lax moder- • Our discourse analysis highlights that QAnon has evolved ation, like Parler (Aliapoulios et al. 2021a) and Gab (Lima into discussing various topics of interest within far-right et al. 2018; Zannettou et al. 2018), or messaging platforms movements across the globe. We find several topics of dis- like Telegram (Dickson 2021). cussions like world politics, conspiracy theories, COVID- Despite that the idea of the QAnon conspiracy theory is 19, and the anti-vaccination movement (RQ3). US-centric, the conspiracy theory recently started becom- Overall, our analysis portraits a nuanced overview of the ing a global phenomenon, and in particular it became popu- QAnon movement across multiple countries. Our observa- lar among people with far-right ideology. During 2020, the tions suggest that the QAnon movement is adapting and QAnon theory spread to Europe (Scott 2020). The conspir- expanding from the US to other countries, usually embod- acy theory is nowadays shared among people from Spain, ied in far-right movements. As a consequence, efforts to re- Italy, United Kingdom, and Germany, one of the most pop- duce redundant work from fact checkers around the world, ular “representatives” in Europe (Bennhold 2020). such as the #CoronaVirusFacts led by the International Fact- Previous work investigates several aspects of the QAnon Checking Network (Poynter 2021), may be relevant to de- conspiracy theory. (Papasavva et al. 2021) analyze content bunk misinformation associated with QAnon globally. toxicity and narratives in a Qanon community on Voat, find- ing that discussions in popular communities on Voat are Background & Related Work more toxic than in QAnon communities.