Source-linked AI summary
Examining the consumption of radical content on YouTube
Homa Hosseinmardi, Amir Ghasemian, Aaron Clauset, Markus Mobius, David M. Rothschild, Duncan J. Watts
TL;DR
The paper asks whether YouTube recommendations and anti-woke channels direct users toward radical political content. Using four years of on- and off-platform browsing data from more than 300,000 Americans, it finds little evidence that YouTube systematically drives radicalization, while anti-woke consumption grows without functioning as a demonstrated gateway to the far right.
Problem
The study addresses limited quantitative evidence for claims that YouTube recommendations or anti-woke channels systematically direct users to radical political content.
Method
The authors analyze individual-level YouTube and off-platform browsing histories from more than 300,000 users who watched over 20 million political-spectrum videos across four years.
Results
Far-right consumption remains small and stable, while anti-woke consumption grows and is linked to far-right consumption off-platform; YouTube recommendations do not systematically explain far-right engagement.
Takeaways & Limitations
YouTube is better understood as part of a larger information ecosystem, and anti-woke content is an increasingly popular category rather than a supported gateway to the far right.
Takeaways & Limitations
Channel categories imperfectly proxy individual-video content, and browsing data cannot reveal recommended videos that users did not choose.
Abstract
from arXiv · showhide
Although it is under-studied relative to other social media platforms, YouTube is arguably the largest and most engaging online media consumption platform in the world. Recently, YouTube's scale has fueled concerns that YouTube users are being radicalized via a combination of biased recommendations and ostensibly apolitical anti-woke channels, both of which have been claimed to direct attention to radical political content. Here we test this hypothesis using a representative panel of more than 300,000 Americans and their individual-level browsing behavior, on and off YouTube, from January 2016 through December 2019. Using a labeled set of political news channels, we find that news consumption on YouTube is dominated by mainstream and largely centrist sources. Consumers of far-right content, while more engaged than average, represent a small and stable percentage of news consumers. However, consumption of anti-woke content, defined in terms of its opposition to progressive intellectual and political agendas, grew steadily in popularity and is correlated with consumption of far-right content off-platform. We find no evidence that engagement with far-right content is caused by YouTube recommendations systematically, nor do we find clear evidence that anti-woke channels serve as a gateway to the far right. Rather, consumption of political content on YouTube appears to reflect individual preferences that extend across the web as a whole.
METHODS AND MATERIALS
The study measures YouTube political-news consumption using a large representative browsing panel, labeled channel categories, missing-label imputation, session construction, and user clustering. These procedures characterize users’ news diets across six political-content categories.
- Data and measurement: N = 309,813 panelists contributed 21,385,962 watched-video pageviews from a nationally representative desktop web panel spanning January 2016 through December 2019.User attention was measured by the duration of in-focus visits to individual videos.
- Video labeling: The researchers mapped prior channel-labeling schemes covering more than 1,100 channels into six categories: far left, left, center, anti-woke, right, and far right.The anti-woke category primarily comprised Intellectual Dark Web and Anti Social Justice Warriors labels, plus a small set of Men’s Right Activists channels.
- Label imputation: 20.1% of video IDs were unavailable through the YouTube API, so missing political labels were imputed using supervised classifiers trained on contextual and user-level features.The classifiers used nearby video categories, viewed news domains, session context, and other individual-level features.
- Label imputation: Random-forest classifiers achieved holdout AUCs of 0.95 for far left, 0.98 for left, 0.95 for center, and 0.97 for anti-woke, right, and far right.Main-text analyses used a high-precision imputation threshold, with an upper-bound model used for robustness checks.
- Session construction: A YouTube session consisted of near-consecutive YouTube pageviews, allowing external pageviews within specified gaps; main analyses used δ = 10 and γ = 60 minutes.The study repeated session-level experiments with alternative gap values to assess robustness.
- User clustering: Users were clustered from normalized monthly viewership vectors across six categories, with hierarchical clustering selecting K = 6 after comparing K values from 2 to 12.Six and five communities received equally strong model-fit support, and six communities were chosen to capture all categories.
RESULTS
YouTube political-news consumption is concentrated in mainstream categories, while far-right and anti-woke audiences remain relatively small but show stronger and stickier engagement. Anti-woke exposure is associated with larger subsequent consumption increases, while cross-platform patterns and referral evidence support user preferences rather than systematic recommendation-driven radicalization.
- News-related content accounts for 11% of total YouTube consumption, with mainstream and generally centrist or left-leaning sources dominating.
- The far-right community remained the second smallest by population at 0.05%, while anti-woke grew considerably and nearly matched center.Far-right population size declined slightly after peaking at the end of 2016.
- 335,209 Americans consumed far-right content and 764,405 consumed anti-woke content in an average month, with anti-woke reach and watch time growing steadily.These results were robust to alternative consumption metrics, inclusion thresholds, and imputation models.
- Far-right and anti-woke communities showed stronger engagement than left, center, and right communities, with anti-woke per-video watch time roughly doubling from 2016 to 2019.Median monthly watch time for far-right and anti-woke users was up to almost twice that of users in the comparison communities.
- Users generally remained in the same political community month to month, indicating that YouTube news-consumption communities were sticky.When users switched communities, movement from the right toward the left was more common than the reverse.
- Bursts of anti-woke content produced larger subsequent-consumption effects than bursts from other political categories, with effects increasing for longer bursts.For a far-right burst of length 4, daily far-right consumption increased by 30.99 ± 19.50 seconds relative to control.
- Anti-woke users were 1.5× as likely to consume right-leaning web content and 2.2× as likely to consume far-right web content as random YouTube users.Right and far-right community members were also more than twice as likely to consume right-leaning content and three times as likely to consume far-right content off-platform.
DISCUSSION
The study finds that YouTube is better understood as part of a broader political-information ecosystem than as a systematic driver of radicalization. Anti-woke consumption is growing and connected to far-right consumption, but the evidence does not support anti-woke channels as a gateway to the far right.
- The study evaluates whether YouTube recommendations and anti-woke channels direct users toward far-right political content.It analyzes more than 300,000 YouTube users and their browsing behavior across a four-year period.
- Far-right content consumption correlates highly with consumption of similar content elsewhere online, indicating that broader news diets matter alongside platform activity.The findings characterize YouTube consumption as reflecting preferences that extend across the web.
- Anti-woke content is growing in both audience size and engagement, while remaining smaller than left-leaning and centrist communities.The anti-woke community also draws members from the far-right more than from any other political community.
- The hypothesized gateway role of anti-woke channels is not supported; anti-woke is instead described as an increasingly popular and sticky category of its own.The paper leaves the broader implications of this growth for future work.
- Most viewing reflects users’ prior tastes and intentions: 80% of sessions last one video, and 55% of videos occur in sessions of no more than four videos.The authors argue that YouTube should be understood as one prominent component of a larger ecosystem for distributing, discovering, and consuming political content.
- The analysis is limited by channel-level content classification, inability to observe recommended-but-unwatched videos, desktop-only browsing, incomplete content coverage, and retrospective labeling.The authors identify these constraints as areas for future methodological improvement.