All of YouTube, Not Just the Algorithm, is a Far-Right Propaganda Machine
YouTube’s celebrity culture and community dynamics play a major role in the amplification of far-right content
In recent years, the media has sounded a constant drumbeat about YouTube: Its recommendation algorithm is radicalizing people.
First articulated by Zeynep Tufekci in a short piece for The New York Times, and later corroborated by ex-Google employee Guillaume Chaslot, the theory goes something as follows: YouTube, in its wish to keep eyeballs glued to its platform, nudges people to more and more extreme content over time. For most types of content, this trend can be harmless, but in the case of political content, it can drive people down “algorithmic rabbit holes” to conspiracy theories or white supremacist propaganda.
This theory has spawned a wave of understandably outraged headlines: “How YouTube Built a Radicalization Machine”; “How YouTube Drives People to the Internet’s Darkest Corners”; “How YouTube Pushes Viewers to Extremism.” In response, YouTube stated last winter that it made changes to its algorithm to decrease recommendations from “borderline” and conspiracy content — although it remained frustratingly vague and opaque about specifics. In fact, YouTube’s lack of transparency has made it nearly impossible to effectively research the algorithm from the outside. Most recently, a study claimed that the algorithm actually led people to less radical and more mainstream content, only to be thoroughly criticized by scholars who claimed the researchers were operating from insufficient data.
So what are we to make of this? Is the YouTube algorithm radicalizing people? Is it just a moral panic generated by an outraged media scared of losing relevance? If it is a problem, can YouTube fix it? Is it even possible for us to know?
I have been researching far-right propaganda on YouTube since 2017, and I have consistently argued that we cannot understand radicalization on the platform by focusing solely on the algorithm. I have also come to find that we don’t actually need to understand the recommendation algorithm to still know that YouTube is an effective source of far-right propaganda. In fact, I will go even further: according to my research, YouTube could remove its recommendation algorithm entirely tomorrow and it would still be one of the largest sources of far-right propaganda and radicalization online.
The actual dynamics of propaganda on the platform are messier and more complicated than a single headline or technological feature can convey — and show how the problems are baked deeply into YouTube’s entire platform and business model. Specifically, when we focus only on the algorithm, we miss two incredibly important aspects of YouTube that play a critical role in far-right propaganda: celebrity culture and community.
For a long time now, YouTube has been a vehicle to stardom. It’s easy to forget that Justin Bieber got “discovered” from his YouTube content, or that Andy Samberg became a household name after his digital shorts went viral there. More recently, when YouTube allowed its users to monetize their content — placing advertisements in front of certain content and giving creators a cut of the revenue — they also incentivized a whole generation of internet users to try their hand at becoming celebrities. Today, the most successful celebrities for a new generation are YouTubers (the numbers are staggering — the most successful individual YouTuber, PewDiePie, has 102 million subscribers at the time of writing).
For years, YouTube has described this in democratizing terms. Indeed, people in their bedrooms can broadcast directly to their fans, creating a sense of intimacy and authenticity not present in older forms of media. In practice, however, that means a range of anti-feminist, Islamophobic, and even white supremacist content creators share far-right propaganda in the form of incredibly intimate personal stories, spoken to their audiences as if they are speaking to close friends. Media historian Fred Turner has described this as a form of “charismatic, personality-centered” authoritarianism. In my own research, I have shown how YouTube creators build trust with their audiences by aligning qualities of authenticity and transparency with reactionary politics.
We don’t actually need to understand the recommendation algorithm to still know that YouTube is an effective source of far-right propaganda
Kevin Roose describes this exact process at work in his incredible New York Times profile of Caleb Cain, a man who was previously radicalized on YouTube before eventually extracting himself. Cain described his own experience as becoming part of a “‘decentralized cult’ of far-right YouTube personalities.” His story centers largely around these personalities, from the white supremacist self-help guru Stefan Molyneux to the young, blonde “identitarian” Lauren Southern, whom he started calling “fashy bae.”
In short, what he describes is a series of parasocial relationships — one-sided relationships in which fans feel as though they genuinely know and are close to the celebrities whose content they view. Parasocial relationships can seem particularly strong when a creator streams for hours on end, and when a viewer, such as Caleb, is lonely or confused. And I argue in my research that it is these relationships — the trust-building, personal storytelling, and seeming authenticity — that convincingly sells audiences on far-right ideas.
But how do people find this content in the first place? Of course, the recommendation algorithm is one answer, but there are multitudes of other ways that far-right content gets disseminated. First, there are other algorithms — the search algorithm and the algorithm that puts content on the home page, for example. None of these technical features exist in a vacuum — influencers explicitly work to maximize their visibility, and far-right influencers have been particularly effective at optimization strategies to appear highly in search results for political terms, for example, using keywords and tagging features.
But people also importantly discover content through something far less technical: social networking between creators and audiences. This can work in a number of ways. Influencers have a direct incentive to collaborate, as it gives them exposure to new audiences and helps provide programming material. These interactions also mean that ideas and viewership can quickly slide between creators, channels, and audiences. When a more extreme creator appears alongside a more mainstream creator, it can amplify their arguments and drive new audiences to their channel (this is particularly helped along when a creator gets an endorsement from an influencer whom audiences trust). Stefan Molyneux, for example, got significant exposure to new audiences through his appearances on the popular channels of Joe Rogan and Dave Rubin.
Importantly, this means the exchange of ideas, and the movement of influential creators, is not just one-way. It doesn’t just drive people to more extremist content; it also amplifies and disseminates xenophobia, sexism, and racism in mainstream discourse. For example, as Madeline Peltz has exhaustively documented, Fox News host Tucker Carlson has frequently promoted, defended, and repeated the talking points of extremist YouTube creators to his nightly audience of millions.
Additionally, my research has indicated that users don’t always just stumble upon more and more extremist content — in fact, audiences often demand this kind of content from their preferred creators. If an already-radicalized audience asks for more radical content from a creator, and that audience is collectively paying the creator through their viewership, creators have an incentive to meet that need. Thus, the incentives of YouTube audiences and creators form a feedback loop that drives more and more extremist content. Influencers are not pure broadcasters, they are part of larger broadcasting communities, and these communities reinforce and spread their ideas to each other and to other audiences and creators.
All of this indicates that metaphor of the “rabbit hole” may itself be misleading: it reinforces the sense that white supremacist and xenophobic ideas live at the fringe, dark corners of YouTube, when in fact they are incredibly popular and espoused by highly visible, well-followed personalities, as well as their audiences. Through parasocial relationships and platform-facilitated social networking, YouTube creators and audiences alike are incentivized to spread and reinforce far-right ideas. In fact, in Mark Bergen’s troubling Bloomberg exposé of YouTube’s corporate culture, he spoke to an employee who had determined that an alt-right “vertical” on the platform received viewership that rivaled the music, sports, and gaming verticals. Thus, YouTube is not just a driver of radicalization; it is a full-fledged far-right propaganda machine.
None of this is to say that the recommendation algorithm doesn’t matter. As a feature responsible for 70% of viewer time on the platform, it clearly plays a crucial role in dynamics on the platform. But it’s just one factor in a broader set of social, economic, and technical issues and incentives baked into the platform. Focusing on the algorithm at the expense of other factors provides a limited view at best and risks minimizing and misrepresenting the problem at worst. To invoke a metaphor from my colleague Whitney Phillips, far-right propaganda on the platform acts more like a pollutant than a rabbit hole: it contaminates those who consume it and simultaneously impacts the whole media environment. The implications of this metaphor are troubling, as they indicate just how big the scope of the problem is. But it is crucial to understand the problem in all its dimensions if we are to take it seriously.