Mapping the Field: Definitions, Scope, and the Online/Offline Divide
Any structure note on this topic must begin with Rothut2026-wt, the scoping review that synthesizes 382 publications spanning 25 years. Its central intervention—that radicalization is an incremental, multi-factor process in which the internet functions as catalyst rather than cause, and that the online/offline binary is no longer analytically tenable—sets the terms for nearly every empirical study filed here. The review’s call for longitudinal, causally-grounded work is directly answered by Bailard2024-pj, which uses Granger causality on 31 months of Proud Boys Telegram data to show that diagnostic and motivational framing (not explicit calls to action) predict subsequent offline violence, tracing a roughly four-week online-messaging-to-offline-action cycle. Together these two papers frame the collection’s central methodological argument: extremism is best studied as a discursive ecosystem unfolding over time, not as a series of discrete flagged posts.
Mainstreaming and Softening: How Extreme Becomes Ordinary
A cluster of papers converges on mainstreaming—the gradual normalization of radical ideas among broader publics—as a process distinct from individual radicalization. Rothut2026-or coins “protest-facilitated mainstreaming” to explain how Germany’s Querdenken movement, though initially grassroots, became structurally embedded with far-right and conspiracist Telegram networks via shared anti-elite framing rather than explicit ideology. Nangle2026-yo extends this logic to Instagram, showing that far-right actors deploy humor, positive affect, and memetic “softening” rather than hostility to normalize extremist worldviews within an “affective-discursive ecology” shaped by algorithmic proximity. Rieder2026-pp offers a parallel case on YouTube: Andrew Tate’s post-deplatforming “ambient ideology” persists and even grows through a decentralized ecosystem of interviews, remixes, and memeified clips that recommendation systems amplify faster than moderators can remove. All three papers converge on a shared claim—that mainstreaming operates through affect, humor, and infrastructural circulation, not argumentative persuasion, complicating content-moderation paradigms premised on identifying discrete “bad” posts.
Platform Affordances and Adversarial Creativity
A second thread interrogates how platform-specific affordances shape extremist tactics. Kakavand2026-kt provides the field’s most systematic comparative account, showing that the same German far-right actors produce structurally distinct networks across five platforms depending on each platform’s relational mix of connectivity, replicability, and scalability—Twitter as an expansive broadcast arena, Instagram as fragmented echo chambers, Facebook as a hierarchical party apparatus. Askanius2026-de complements this with a case study of Sweden’s YouTube “alternative influence network,” showing how influencers fuse microcelebrity branding and monetization with propaganda, using YouTube as a “sanitised” gateway funneling audiences toward alt-tech platforms via coded language and phantom-video workarounds. Karo2026-dn pushes the adversarial-creativity argument furthest, cataloguing five concrete evasion tactics—audio camouflage, meme infiltration, blurred intent, emoji codes, bait-and-switch—used by jihadist supporters on TikTok, arguing that this “everyday extremism” exploits structural moderation flaws that are not ideology-specific but shared across far-right, anti-vaccine, and jihadist communities. Hollingshead2026-vx complicates the picture by showing that on TikTok, at least in the Canadian immigration debate, popular content skews pro-immigration and that both sides deploy the platform’s affordances in symmetrical ways—suggesting platform vernacular is ideologically agnostic even as it can also enable racist “speculative worldbuilding” via generative AI.
Epistemics of Belief: Redpilling, Conspiracy, and Boundary Work
Where the previous clusters emphasize infrastructure, a set of papers turns to the epistemic and identity work that sustains belief. Marwick2025-vx dismantles the “exposure equals belief” model of radicalization, showing through redpilling narratives on Reddit, Gab, and Discord that adoption of far-right belief is typically a processual socialization in which disinformation is marshaled as “evidence” to construct a self-image of rational, scientific superiority. Marwick2025-ov extends this epistemic lens to ConspiracyTok, revealing a demographically unexpected cohort of young, non-white, and female creators who build “generous epistemology” and populist knowledge production that can simultaneously empower marginalized standpoints and circulate antisemitic or transphobic disinformation. Grusauskaite2026-po generalizes this insight across six platforms in its study of “Great Reset” discourse, arguing explicitly against techno-centric echo-chamber explanations: conspiracy culture is sustained by symbolic boundary work—reclaiming “conspiracy theorist” as a badge of honor, converting personal hardship into affective solidarity—that unites ideologically disparate groups (secular-libertarian and religious-Christian) around a shared anti-institutional “we.” These three papers collectively insist that content and platform affordances alone cannot explain radicalization; identity, affect, and self-legitimizing epistemics are the deeper mechanism.
Coordination, Manipulation, and the Limits of Detection
A final cluster shifts from organic extremist communities to engineered manipulation. Bastos2025-ol mines the Twitter Moderation Research Consortium dataset to show that state-linked troll farms exhibit detectable visual signatures in profile imagery, extending detection research beyond text and network signals. Jovanovic-Harrington2026-ze and Kulichkina2026-zk both examine coordinated inauthentic behavior in authoritarian contexts—Serbia’s SNS party and Chinese state-aligned accounts during COVID-19 protests, respectively—showing that manipulation is evolving toward subtler, harder-to-detect mimicry of authentic grassroots activity, and that social media functions as dual-use infrastructure for both mobilization and repression. Simeone2025-vo closes this thread with a rare demonstration that targeted deplatforming can work: removing a handful of central Arizona Election Review accounts collapsed the right-leaning cluster’s hub structure and shifted discourse from mobilization to meta-conspiracy about the bans themselves, a finding that stands in productive tension with Rieder2026-pp’s account of Tate’s moderation-resistant ecosystem—suggesting that the efficacy of platform intervention may hinge on network centralization versus diffuse, remix-driven circulation.
Synthesis
Read together, these sixteen papers trace an arc from mapping (what radicalization research has established, per Rothut2026-wt) to mechanism (how online discourse translates into offline action, per Bailard2024-pj) to infrastructure (how platform affordances are exploited and adapted, per Kakavand2026-kt, Karo2026-dn, Askanius2026-de) to culture (how affect, humor, and epistemics sustain belief and belonging, per Nangle2026-yo, Marwick2025-ov, Marwick2025-vx, Grusauskaite2026-po) to governance (the contested efficacy of moderation and detection, per Rieder2026-pp, Simeone2025-vo, Bastos2025-ol, Jovanovic-Harrington2026-ze, Kulichkina2026-zk). The through-line uniting them is a shared rejection of simple, content-centric models of radicalization in favor of accounts that foreground temporality, affect, platform-specific sociotechnical relations, and identity-based meaning-making—an agenda Hollingshead2026-vx’s finding of ideologically symmetrical platform vernacular usage crystallizes: the danger of extremist normalization lies less in any single piece of content than in the shared infrastructures, genres, and emotional registers that make radical and mainstream discourse increasingly difficult to tell apart.