�鶹������ý

There are a couple ways to think about how online content might have influenced the outbreak of violence in Washington on January 6. One is to treat the siege as a one-off event—an aberrant security breach that is unlikely to happen again. Another is to view the attack as the culmination of a years-long influence campaign involving some of �鶹������ý most powerful allies. Either way, the clear aim of the many thousands of Americans who stormed Congress that day was to halt the certification of the vote count in the 2020 presidential election and prevent the peaceful transfer of power to �鶹������ý political challenger, Joe Biden.

We know this because thousands who took part in the violence that day stated as much on platforms like Parler, Twitter, and Facebook. Much of what we know about what happened on January 6, in fact, comes from a virtual volunteer army of digital sleuths, data nerds, hackers, and investigative journalists who raced to capture as many social media posts as possible before the evidence of the violence disappeared. Indeed, FBI Director Christopher Wray estimates that the agency received upwards of 200,000 digital media tips about the events of January 6.¹

On Twitter, in the immediate aftermath of the insurrection in Washington, footage of a man in a knit cap bearing the logo of the Chicago Fire Department wielding an extinguisher near the steps where Capitol Police Officer Brian Sicknick was killed went viral under the hashtag #extinguisherman.² On Facebook, references to the #StopTheSteal campaign went viral. Parler, Twitter’s self-described conservative rival, exploded with posts thrown up by packs of Proud Boys, Oath Keepers, Three Percenters, and other white-supremacist militias who descended on Washington in the weeks leading up to January 6.³ The deaths of two pro-Trump protesters, Ashli Babbitt and Roseanne Boyland, at the scene were recorded and uploaded online on multiple platforms for all the world to witness, and have since become a cause ��é��è������ for Trump and causus belli for far-right extremists.⁴

With roughly 15 million users at the time of the attack, Parler’s user base paled in comparison to the billions who use Facebook and hundreds of millions who use Twitter.⁵ Indeed, Parler then and now represents only a small corner in an expanding constellation of social media platforms with loose content moderation policies that have cropped up as alternatives as big tech firms have stepped up their moderation of content in recent years. These so-called “alt-tech” sites began to proliferate and take on more prominence as Trump began to prepare his re-election campaign amid a series of White House scandals. But, because so much of what the world witnessed on January 6 was culled from Parler posts and authorities have publicly linked Parler to planning of the attack, the platform has emerged as the poster child for unbridled online extremism in the United States.⁶

When Parler first launched a little more than three years ago, it billed itself as a free speech, censorship-free alternative to Twitter and other mainstream platforms.⁷ In the wake of the January 6 attack, Parler executives have, however, argued that the company is no different from mainstream platforms. In a March 2021 letter to the House Committee on Oversight and Reform, in fact, the Nevada-based social media company argued that it proactively sought to warn the FBI about concerning content posted on its platform before Trump appeared on stage at the “March to Save America” rally on the White House ellipse on January 6.⁸

Parler’s attorneys, in fact, asserted in the same letter that big tech companies were unfairly trying to scapegoat its platform for fomenting the violence at the Capitol, and suggested that big players like Facebook and Twitter were more culpable in allowing false and hate-filled content to circulate widely on their sites before the mob siege in Washington.⁹ In lawsuits and public statements, Parler’s stakeholders have continued to deflect blame for the mob assault onto other platforms, arguing that Amazon and other big tech companies singled Parler out because they wanted to crush market competition.¹⁰

Although Parler’s reach appeared to be smaller than its competitors, Amazon's move to deplatform the social media site on January 10, four days after the Capitol attack, sent enormous shockwaves through the tech industry. It also drew outsized attention from the media, advocacy groups, and Congress, prompting new calls for revisions to Section 230 of the Communications Decency Act,¹¹ the regulation that gives Internet platforms immunity from adverse legal action related to what their users post online.¹²

For the above reasons and more, evidence culled from Parler and other fringe social media sites will be pivotal for the House Select Committee’s inquiry.¹³ As of late November 2021, the Select Committee had issued subpoenas to key organizers of the January 6 rally on the White House ellipse that lit the fuse for the violence at the Capitol.¹⁴ While some key witnesses are resisting the panel of inquiry’s requests for information and Trump himself is locked in a legal battle over potentially relevant records, it is expected that the 35 technology companies that have received requests for information to date will be relatively compliant.¹⁵

Accordingly, the Select Committee has also issued demands to Parler and other tech companies to preserve and produce records pertaining to the spread of misinformation and disinformation about the 2020 presidential elections, efforts to overturn or prevent the certification of election results, and foreign influence in the election. Of particular concern for the committee and the public are actions and policies undertaken by tech companies to manage the avalanche of false information about the elections and to manage the adverse impacts of malign influence campaigns.¹⁶

The House Select Committee has expressly said that its investigation will examine “influencing factors that contributed to the domestic terrorist attack on the Capitol and how technology, including online platforms, financing, and malign foreign influence operations and campaigns may have factored into the motivation, organization, and execution of the domestic terrorist attack on the Capitol.” This is a good start, but with dozens of social media and tech companies implicated in the January 6 attack, there is a lot of ground to cover.

What we know already is that the de-platforming of Parler and tech industry sweeps triggered and accelerated a virtual stampede to lesser-known alt-tech platforms like Gab, MeWe, and Rumble.¹⁷ That became especially evident after an anonymous hacker leaked a massive 70-terabyte cache of Parler data containing millions of digital hints about the scope of the coordinated messaging and planning that went into �鶹������ý online influence campaign.¹⁸ The Parler data haul has been the subject of countless media exposés, and has attracted the attention of journalists and researchers around the world.¹⁹

Yet, questions persist about how to interpret millions of posts on Parler and other platforms, what the data reveals about the events of January 6, and how social media content moderation policies might impact future elections and political violence in the United States. Here are just a few questions that we set out to answer with the launch of this project:

• Was there something unusual about the way Parler operated that made its networked effects on the information ecosystem different from those of its competitors?
• Was there evidence of any online coordination between influential Trump campaign insiders and surrogates and far-right online influencers in amplifying false content?
• How much did the rise of so-called alt-tech media, online hubs that sprang out of far-right extremist, white supremacist, and counterfactual communities, contribute to the violence on January 6?
• Is there any way to accurately assess competing claims made by big tech companies like Amazon, Facebook, and Twitter about whether fringe platforms like Parler were more culpable in fomenting online hate, mis- and disinformation about the elections, or polarizing issues like policing and COVID-19 restrictions?
• Was the riot partially a result of a coordinated strategic influence campaign where non-state actors or quasi-governmental proxies like the Internet Research Agency in Russia and state intelligence agencies like Russia’s Main Intelligence Directorate (GRU) and/or China’s People’s Liberation Army (PLA) collaborated in promoting polarizing content online?

Almost from the instant violence began to erupt at the Capitol on January 6, Parler emerged as the central focus of a loosely organized network of volunteer researchers who banded together across several time zones to monitor social media. Bellingcat, a nonprofit investigative research collective, was among one of the leading organizations who spearheaded a call to collect and archive social media data.²⁰ Several on our team answered that call.

In the process, we collected and archived the accounts for 100-plus Parler users with close ties to the Trump campaign. That subset of accounts contains data about the online behavior of key influencers such as Sidney Powell, a one-time Trump campaign lawyer; �鶹������ý former national security adviser Michael Flynn; former New York City Mayor and Trump campaign lawyer Rudy Giuliani; and several dozen other lesser-known influencers and adherents of the Oath Keepers, Three-Percenters, and Proud Boys movements. We continued to collect from a variety of additional sources, including Gab, BitChute, Telegram, and Rumble, among others. But collection on Parler ended on January 10, 2021 when Amazon took Parler offline and Apple and Google stopped hosting Parler’s mobile app.²¹

Unbeknownst to much of the world at the time, however, a social media analyst and self-proclaimed “hacktivist” who uses the Twitter handle @donk_enby had already leapt into action well before the platform’s shutdown, archiving as much of the Parler data as possible.²² The hacktivist, who had already published the results of a study into the unofficial Parler API,²³ downloaded, archived, and made publicly available a million videos from an unsecured Parler S3 bucket on Amazon’s cloud storage service.²⁴ This data, too, became part of our collection.

Faced with the daunting task of trying to make sense of digital data documenting the violence and of tens of millions of social media posts and tens of terabytes of raw data, we knew we had a very hard puzzle on our hands. With more than 500 people under indictment when we started our investigation in June 2021 and only sparse information about whether and how social media data might be relevant in understanding their journey to Washington, the first challenge was to decide what exactly to focus on.

We opted, therefore, to focus first on the “StopTheSteal” movement itself and Parler as a platform second. If we approach Parler as “preparatory media,”²⁵ that is, as media scholar Luke Munn put it, the venue that delivered the “logistical, organizational, ideological” elements to turn a mob into a cohesive force for political violence, then what did the user data look like in the events leading up to and on the day the violence unfolded in time and space? To find out, we applied a mix of methods, including open-source investigation, natural language processing, machine learning, and social network analysis to map demographic and other trends in publicly available data.

In the lead up to January 6, the beating heart of the “Stop The Steal” movement—online and offline—was the interplay between three slices of the demographic that supported Donald �鶹������ý false claims about election fraud. Besides Trump himself, the first among them were the influencers, high-profile insiders close to the Trump campaign and White House who publicly promoted falsehoods about voting irregularities like Powell, Giuliani, Flynn, My Pillow CEO Mike Lindell, and Trump campaign adviser Roger Stone.²⁶ Then there were the objectors, 147 members of Congress who formally lodged objections to the certification of the elections and voted to overturn the results.²⁷

Last, but by no means least, there were the hundreds of Americans who answered the call of political elites on Facebook, Twitter, Parler, Telegram, Gab, thedonald.win, 8kun, and other platforms. Although not all expected to join in the violent putsch aimed at contesting the election results, many did and their efforts to contest the outcome shocked the world. All three categories of social media users—influencers, objectors, and contesters—are also at the center of the inquiry behind this report.

The analysis in this report reflects our first attempt to seek answers about how Parler and other platforms factored into what happened online and what happened offline, in the real world. Based, in part, on an early assessment of a cache of an estimated 183 million Parler posts publicly archived after it was temporarily deplatformed, the analysis in this report offers unique insights into trendlines that emerged over the period spanning January 2020 to January 2021. Our analysis also relies on unique location-based event datasets, publicly available information about individuals charged in connection with the attack, and open-source data culled and curated from social media sites, public records, media accounts, and academic and think tank literature.

In all, our team has collected and begun to analyze over a million video posts, over a million image posts, over 183 million posts, and metadata from 13 million plus user accounts culled from publicly available Parler data. Additionally, we collected Twitter data from the 147 members of Congress who formally objected to the certification of the election results; and we analyzed the Parler user accounts for close to a quarter of those officials. We have also collected and curated unique data relevant to the January 6 attacks and worked with partner organizations to understand what all this data tells us about the interplay between online discourse and real-world events such as demonstrations, protests, and political violence of the kind witnessed on January 6.

The report is divided into seven sections, including this one, which provides context about the attack and a brief explanation of our research methodology:

• The second section examines the evolution of Parler’s rise from a startup social media provider with a relatively small online following to one of the topmost downloaded mobile phone apps just ahead of the 2020 election before it was taken temporarily offline in January 2021.²⁸ The second section also outlines what is known about publicly available Parler data from the 1.0 version of the site.
• In the third section, we detail initial observations with a look at those indicted in connection with January 6 and the trajectory of specific individuals with Parler accounts who played a direct role in stoking the “Stop the Steal” movement. In this section, we also explore the connections between real world political events and demonstration activity and the scores of Americans who showed up in Washington to contest the election results.
• In the fourth section, we look at the impact of elite signaling from prominent personalities and elected officials through an assessment of the Parler accounts for a subset of influencers close to the Trump campaign and militia groups. This section includes a topic modeling and link analysis for this slice of Parler accounts that reveals the outsized impact that being banned from Twitter, Facebook, and other sites had on Parler.
• In the fifth section, we examine the social media profiles of the 147 members of Congress who formally raised objections to the certification of the Electoral College count and voted to overturn the election.
• The sixth section summarizes several observations about user behavior on Parler culled from video posts on the platform.
• The seventh and final section describes potential implications and outlines possible new lines of inquiry that may inform future research.

In line with all the above, the Future Frontlines team and its partners set out in June 2021 to learn more about the relationship between online behavior on Parler and other social media platforms and offline behavior in the year leading up to the Capitol attack. Our research questions appeared deceptively simple: How did Parler factor into the “Stop The Steal” movement? Was there a way to tell whether and how Parler played a role in mobilizing �鶹������ý supporters to violence?

Relatedly, one additional question was: How could we make sense of Parler’s claims that its role in mobilizing people to violence on January 6 was no more significant than that of its rivals Twitter and Facebook? In other words, is there a way to assess the impact of fringe alt-tech social media sites that market themselves as champions of “censorship free” expression on social movements and mobilization to violence? How did interactions between Parler users and users on other platforms like Twitter come into play?

To answer those questions, we needed to know more generally about Parler’s overall operational model and how the company’s guiding philosophy mapped to the platform’s design features and operations. We also needed to understand what the publicly available data from the 1.0 version of the platform reveals about Parler’s user base and design features. We needed to know how Parler’s user base engaged on other social media platforms that were identified in federal charging documents and congressional reports and requests for information relevant to the events of January 6. We also needed to learn more about how offline events such as protests, rallies, and major news developments intersected with online activity.

Our joint team of investigators mined the data over a period of roughly six months and deployed a variety of tools to elicit a set of baseline observations, including computational analysis on metadata contained in the files, topic modeling of posts, and open-source investigation techniques. We also compiled two unique datasets.

One dataset consists of publicly available Parler posts of prominent and prolific Parler users, selected for their influence on the platform as determined by either their close ties to the Trump White House or President �鶹������ý presidential campaign, or links to militia groups such as the Proud Boys, Oath Keepers, and Three-Percenters. This specially curated dataset of Parler influencers contains approximately 40,000 posts, with about 30,000 containing text. While this sample cannot be categorized as representative because it is small relative to the roughly 183 million Parler posts in publicly available datasets, it does offer a window into how key Trump supporters framed their concerns about the elections and other polarizing causes that roiled the nation in the lead up to January 6.

Another dataset focused on the scores of Americans who were arrested after they answered �鶹������ý call to publicly contest the elections. Collected from publicly available charging documents issued by federal prosecutors, this dataset includes information about 632 people arrested in connection with the Capitol attack from early January to the end of October 2021. Unlike comparable datasets, our list of arrestees references contemporary news reports to trace individuals back to their location of residence, rather than relying on location of arrest as a surrogate for residence. This allows for more fine-grained analysis of rioters' origins, as arrests often took place some distance from arrestees’ homes.

To contextualize larger real-world trends, we also tapped into data from the U.S. Crisis Monitor supported by our partners at the Bridging Divides Initiative at Princeton University through their partnership with the Armed Conflict Location and Event Data Project (ACLED).²⁹ The Bridging Divides Initiative (BDI) supports data-driven initiatives to build community resilience and facilitate efforts to track, prevent, and mitigate political violence in the United States. The U.S. Crisis Monitor project also provides in-depth insights into the potential risk for the escalation of political violence and identifies opportunities for de-escalation in analytical briefs. Since 2020, ACLED has systematically collected the dates, actors, locations, fatalities, and types of all political violence and demonstration events across the country.³⁰ That data provides granularity on a number of trends mirrored in the online data we analyzed.

One U.S. Crisis Monitor analytical brief in particular, published in December 2020, stands out for its assessment of the potential for right-wing violence during the delicate period after Election Day on November 3 and leading up to the inauguration on January 20, 2021. The ACLED brief notes an uptick in right-wing events, which it defines as being motivated by “distinct” but interrelated causes, “including: pro-Trump or anti-Biden rallies,” pro-Republican Party, pro-police or “Blue Lives Matter,” anti-Black Lives Matter protests, and “demonstrations that involve QAnon conspiracy theory supporters or protesters associated with the ‘Save Our Children’ movement,” and protests involving right-wing militias or “street movements.”³¹

The brief noted that right-wing groups largely mobilized over the summer of 2020 in response to Black Lives Matter (BLM) demonstrations and public health related responses to COVID-19. While some militia groups like the Oath Keepers and street movements like the Proud Boys suggested poll monitoring on the day, it wasn’t until after Trump tweeted false claims about the outcome on November 4 that there was a notable increase in “Stop The Steal” demonstrations. We wanted to see if there was a similarly notable uptick of Parler posts referencing key right-wing themes like “Stop The Steal,” “Back the Blue,” and left-leaning counterprotest themes like “Count Every Vote.” We also wanted to see what, if any, spatial or temporal patterns could be discerned from the Parler data. To do this, we deployed topic modeling tools and analyzed geocoded metadata from video posts on the platform.

Our cursory assessment found Parler 1.0 seemed to gain more engagement over time as protests and demonstrations in response to public concerns about policing practices, COVID-19-related restrictions, and the 2020 election were mirrored in online discourses on mainstream platforms. User engagement on Parler, indeed, appeared to grow almost in tandem with protest events related to the election as the November 3 deadline for voting neared. However, it is worth repeating the old saw that correlation is not causation; the leaked data culled from Parler 1.0 has several significant limitations, and a lot more contextual data is needed to understand what, if any, relationship there is between online engagement on Parler and offline events.

It is important to emphasize that our observations are only preliminary and pertain only to publicly available data from Parler 1.0, the version of the platform as it existed before Amazon took it offline in January 2021.³² We lay out the weaknesses in the publicly available data from Parler 1.0 in more detail further below. But it is important to state up front one notable limitation of the data is that only so much is known about the techniques used to collect and/or curate Parler post data from the early version of the platform. Moreover, it cannot be said that the data represents the entirety of Parler’s user base pre-January 10, 2021. The picture this data delivers is at best richly pixelated and only represents a moment in time. Still, there is much that can be learned from examining the available data culled from some 4 million plus users in one set and 13 million users in another.