The Platform Infrastructure Behind Generational Antisemitism
A Response to Yair Rosenberg’s Atlantic Article
AddressHate Research Scholar at NYU’s Center for the Study of Antisemitism | Postdoc, University of Cambridge | Lead, Decoding Antisemitism | Research Advisor, AddressHate
Yair Rosenberg’s recent Atlantic piece documents a striking pattern: antisemitism in the United States is no longer primarily a partisan phenomenon, but a generational one. Drawing on multiple large-scale surveys, Rosenberg demonstrates that twenty-five percent of Americans under 25 hold unfavorable opinions of Jewish people, with negligible differences between Trump and Harris supporters. Age predicts antisemitic attitudes better than political affiliation—a pattern that cannot be explained by partisan alignment alone.
This finding is empirically robust—and analytically incomplete. It contradicts the comfortable assumption that bigotry naturally fades with demographic change. But to understand why this generational divide exists requires examining the infrastructures of social media platforms and the communication conditions through which political meaning is now formed.
Rosenberg’s analysis identifies structural factors rather than partisan blame. He points to how social media platforms with weak moderation allow conspiracy theories to spread rapidly, algorithmically amplified because they generate engagement. He describes these platforms as an “escalator” that gradually funnels users from legitimate criticism toward explicitly antisemitic material. Crucially, he emphasizes that antisemitism remains a minority position—most young Americans reject it, but its visibility is amplified by outrage-driven media dynamics.
This framing validates what years of analyzing digital discourse have been demonstrating about how antisemitism actually spreads. The generational pattern isn’t about young people being uniquely susceptible to prejudice—it’s about the information infrastructure through which they form political understanding. What makes the convergence between Rosenberg’s work and ours particularly significant is the complementarity: his survey data provide population-level validation of patterns we’ve been documenting at the discourse level. Where our analysis shows how antisemitic content circulates through platform mechanisms, his surveys demonstrate that these mechanisms are having measurable effects on attitudes across entire demographic cohorts.
The statistics Rosenberg presents demonstrate why social media research is essential to understanding contemporary antisemitism. When survey data show dramatic age-based differences but no partisan gap, that reveals where political mindsets are actually being formed. For young people, social media platforms have become the primary arena for political formation—the space where frameworks for understanding complex issues develop. If we want to understand why 25% of Americans under 25 hold unfavorable views of Jews, we need to study the platforms where they encounter political information.
This response makes three interconnected arguments. First, it shows how the Decoding Antisemitism project’s discourse analysis contributes a mechanistic explanation for Rosenberg’s population-level findings—documenting how antisemitic content actually circulates across platforms and ideological contexts. Second, it translates Rosenberg’s “algorithmic escalator” metaphor into a concrete analysis of implicit coding, content moderation gaps, and cross-platform diffusion. Third, it proposes structural interventions that address how antisemitism spreads rather than attributing it to political camps or demographic groups. The goal is shifting focus from who holds antisemitic attitudes to the infrastructure that produces and amplifies them.
What the Research Shows About Cross-Spectrum Dynamics
The Decoding Antisemitism project, which has analyzed over 300,000 pieces of digital content across European and American contexts since 2020, offers a mechanistic account of what Rosenberg’s survey data reveal at the population level. Our work began by examining how antisemitic discourse manifests in mainstream comment sections across multiple countries and platforms. What emerged was a pattern that should fundamentally reshape how we think about contemporary antisemitism.
Different political contexts have different ways to communicate antisemitic ideas. This holds true of progressive, liberal, and conservative online spaces. The specific tropes vary—conspiracy theories about global elites in some contexts, accusations of colonial oppression and fascist crimes in others, claims about dual loyalty and classic tropes of malevolence, hypocrisy, and greed in still others. But the underlying mechanisms through which these ideas spread remain remarkably consistent across ideological boundaries.
Our most recent analyses examined comment sections following antisemitic attacks. In the Bondi Beach shooting case, we analyzed YouTube coverage from five major American outlets. Moving from left to right across the political spectrum, CNN’s coverage produced the highest rates of antisemitic discourse at 17%, ABC at 6%, NBC at 5%, Forbes at 14%, and Fox News at 10%. Notably, the Bondi case revealed something that challenges simplistic assumptions: Israel-related antisemitic tropes—false flag theories invoking Mossad, accusations that Jewish orchestration of immigration policies led to the attack, inversions blaming Israeli actions for antisemitic terror—appeared across the political spectrum, including in conservative-leaning comment sections. The patterns we documented included Great Replacement narratives explicitly attributing demographic change to “Jewish globalists,” conspiracy theories about Israeli intelligence staging attacks, and justifications framing the massacre as karmic consequence of Gaza.
For the Washington Embassy shooting in May 2025, where two Israeli diplomatic staff were murdered, we examined coverage from eight outlets spanning the Anglophone media landscape. Arranged from left to right: Al Jazeera English (66% antisemitic discourse), Guardian (40%), BBC (42%), CTV News (50%), The Times (54%), Forbes (28%), Sky News (40%), and Fox LiveNow (24%). The Washington case showed particularly strong cross-spectrum presence of conspiracy theories, with “false flag” narratives, mockery of Jewish grief (”kvetching intensifies”), and justifications invoking Gaza appearing across outlets regardless of ideological orientation.
These findings build on DA’s six Discourse Reports, which examine antisemitic responses across Germany, France, and the United Kingdom, utilizing different discourse events, languages, and online spaces. That larger corpus analyzed diverse trigger events, including antisemitic incidents, Holocaust commemorations, and Middle East escalation phases, through rigorous critical discourse analysis. The pattern held consistently: each type of event generated antisemitic outrage in social media comment sections of progressive, liberal, and conservative media outlets alike.
The pattern that emerges isn’t about left versus right. It’s about how discourse events trigger antisemitic responses across different media ecosystems, with variation correlating more strongly to platform characteristics, moderation standards, and event framing than to outlet ideology. Israel-related antisemitism—once primarily associated with left-progressive spaces—now circulates widely in right-wing contexts as well, often merging with older conspiracy frameworks about Jewish power and civilizational decline.
Documenting prevalence across aligned audiences is not a claim of moral symmetry but of infrastructural vulnerability. The mechanisms through which antisemitic content evades moderation, gets algorithmically amplified, and spreads across platforms operate regardless of the ideological framing that activates them. This is why the cross-spectrum presence matters: it redirects attention from political blame games toward the platform architectures that enable spread.
Of course, everything related to October 7, Israel, and the Middle East conflict has created asymmetric effects. There is substantial demonization of Israel and glorification of terror, primarily in left-progressive spaces—the inversions casting Israeli self-defense as genocide, the justifications treating Hamas’s massacre as resistance, the systematic denial of Jewish trauma. But we cannot ignore the staggering numbers of antisemitic discourses we document in conservative contexts as well. The cross-spectrum presence is an empirical reality, even as its specific manifestations and intensity vary across different political contexts. Different tropes, different triggers, different justification strategies—but the underlying platform mechanics that allow implicit content to evade moderation and reach mass audiences operate across ideological boundaries.
These findings describe the prevalence of antisemitic discourse within aligned outlet audiences during specific discourse events. They don’t imply that antisemitism is caused by political orientation. What they demonstrate instead aligns with Rosenberg’s generational finding: if age predicts antisemitism better than politics, this points toward platform infrastructure and event dynamics as explanatory factors rather than partisan identity. When we analyze comment sections under ideologically legible outlets, we’re examining where audiences aggregate, not why antisemitism appears—outlet ideology shapes which audiences gather, but platform affordances shape how antisemitic discourse circulates once they do.
The Algorithmic Escalator and Platform-Mediated Conditions
When Rosenberg describes social media as an “algorithmic escalator” that funnels users from legitimate criticism toward antisemitic radicalization, he’s identifying something our research has documented mechanistically. But the escalator metaphor, while apt, risks oversimplifying the problem as purely algorithmic. What actually operates is a fuller set of platform-mediated communication conditions.
Our DA taxonomy of over 170 classification units for hateful discourse reveals how contemporary antisemitism reaches young people through both explicit slogans and coded patterns. The survey data Rosenberg cites don’t distinguish between these forms. But what makes the generational pattern so pronounced is how implicit, coded discourse provides the scalable pathway through which antisemitic frameworks become thinkable, defensible, and politically legible over time.
Coded language activates established tropes without explicit slurs: “Certain people seem to have outsized influence in media and finance” doesn’t mention Jews but triggers the same conspiratorial framework; “dual loyalty” accusations function as apparently political criticism while drawing on centuries-old patterns of Jewish othering. This implicit discourse isn’t the only form antisemitism takes online—explicit hate remains visible and often sanctioned—but implicit content is scalable, deniable, and cumulative in ways explicit hatred cannot be.
The platform-mediated conditions enabling this normalization include but extend beyond algorithms: algorithmic recommendation systems surface progressively extreme content because engagement metrics reward conflict and conspiracy; anonymity structures reduce social accountability; participatory comment cultures create mutual reinforcement dynamics where extremity becomes normalized through repeated exposure; the omnipresence of hate speech renders it ambient rather than exceptional; collapsing contextual boundaries blur news, memes, and commentary; rapid, highly emotional interaction discourages reflective evaluation; strategic ambiguity and plausible deniability allow antisemitic tropes to circulate as apparently legitimate political analysis, with creators claiming “I didn’t say Jews, I said globalists” even as audiences receive the intended message.
These conditions create environments where minority hate becomes majority-visible discourse through platform mechanics rather than demographic preference. Tim Miller’s observation—the host of the Bulwark Podcast—about students “starting to think that the Jews killed Charlie Kirk” captures the endpoint: conspiracy thinking that seems organic but results from systematic exposure to implicit framings amplified through engagement-optimizing systems.
Why Cross-Spectrum Documentation Matters
When Guardian and Sky News show comparable prevalence during the same discourse event, that tells us outlet ideology isn’t the primary driver. When variation across American outlets doesn’t track with political orientation—NBC at 5%, ABC at 6%, Fox at 10%, Forbes at 14%, CNN at 17%—that contradicts simplistic partisan explanations. This cross-spectrum presence isn’t about claiming equivalence. The manifestations differ significantly: Israel-related events trigger different antisemitic tropes than conspiracy-focused events. But the underlying mechanisms—implicit coding, algorithmic amplification, event-driven activation—operate across ideological contexts, redirecting attention from political blame games toward structural intervention points.
Event-Driven Dynamics
Israel-related conflicts trigger massive spikes in antisemitic discourse across all outlets. Financial crises activate conspiracy theories about elites and globalists. Terrorist attacks produce false flag accusations and dual loyalty tropes. The content clusters around discourse events that provide legitimizing frames for antisemitic patterns.
Young people consume content through engagement-optimized platforms where each major news event becomes an opportunity for antisemitic discourse to spread. Rosenberg identifies the loss of Holocaust memory as one contributing factor. Our research suggests this historical amnesia combines catastrophically with platform dynamics: when traditional guardrails weaken simultaneously with platforms that algorithmically amplify extreme content, antisemitism finds new pathways providing “cognitively cheap answers to complex crises.”
The Structural Reality
Both Rosenberg’s survey data and our mixed-methods discourse analysis converge on a crucial insight: antisemitism in 2025 is fundamentally a structural problem. It travels through platform architecture, not demographic essence. It activates through discourse events, not partisan ideology. It persists through implicit coding that evades traditional moderation.
Rosenberg’s framing deliberately avoids partisan blame—he shows that age matters more than political affiliation, that the problem crosses ideological boundaries, and that solutions require consensus-building rather than political victory. Our research provides the mechanistic validation for this approach. The similar rates across Guardian and Sky News audiences during discourse events tell us where to look for solutions: not at which political camp to defeat, but at which platform mechanisms to reform.
Yet the consensus Rosenberg identifies—57% of young people rejecting antisemitism—fails to mobilize effectively. Understanding why requires recognizing how antisemitism is routinely externalized along three axes, each shifting attention away from the structures that actually enable its spread.
Externalization operates outward to foreign actors—blaming bots, hostile states, or external influence operations while avoiding examination of domestic platform architecture and homegrown participatory dynamics (see also my article on X’s location feature). It operates downward to fringe extremists—framing antisemitism as marginal, exceptional, and unrepresentative while obscuring normalization in mainstream comment sections and diffusion through influencer networks. And crucially, it operates sideways to the opposing political camp—turning antisemitism into evidence of the other side’s moral failure while immunizing one’s own camp against self-examination.
This sideways externalization is perhaps the most socially corrosive because it aligns perfectly with platform incentives. Demonizing the other political side generates engagement, rewards outrage over introspection, and hardens camps rather than addressing mechanisms (see also our study on the digital aftermath of the Charlie Kirk murder). Left spaces minimize antisemitism when it appears as “anti-Zionism,” right spaces minimize antisemitism when it appears as “anti-globalism,” and both focus obsessively on the other’s worst examples. Platforms amplify this dynamic because conflict between camps drives metrics.
Antisemitism, as a hyper-adaptable hate ideology, is effectively outpacing the oversimplified binary frameworks through which public discourse is organized. While political camps argue about whether antisemitism is fundamentally a left or right phenomenon, antisemitism itself morphs across ideological contexts, exploiting whatever rhetorical opportunities each political space provides. The dichotomic framework isn’t just inadequate—it actively enables antisemitism’s spread by preventing recognition of how it operates across boundaries.
The interpretation of antisemitism is itself polarized, and that polarization becomes a vehicle for antisemitism’s persistence. This explains why Rosenberg’s finding matters so profoundly. The generational gap with no partisan gap only makes sense if antisemitism is being reproduced through shared communication environments while political camps are busy externalizing blame sideways. If antisemitism were foreign manipulation, age wouldn’t predict it; if it were fringe extremism, prevalence would be low everywhere; if it were genuinely partisan, we’d see the left-right split that’s conspicuously absent. Rosenberg’s data pattern points directly toward platform-mediated conditions as the explanatory factor—and toward sideways externalization as the reason consensus fails to translate into action.
This is simultaneously more depressing and more hopeful than partisan narratives allow. More depressing because the problem is structural rather than confined to one political camp, which could theoretically be defeated. More hopeful because structural problems admit structural solutions—interventions that don’t require changing hearts and minds one person at a time, but rather changing the systems that shape how information flows. These are the mechanisms through which Rosenberg’s “escalator” operates, and these are the points where intervention becomes possible.
From Mechanism to Intervention
Understanding how antisemitism spreads reveals where intervention points exist. If implicit antisemitism evades keyword filters, detection systems need training on actual discourse patterns. If algorithmic amplification surfaces extreme content for engagement, transparency requirements and deamplification protocols become essential. If cross-platform migration spreads coded language faster than moderation adapts, industry-wide coordination matters more than individual platform policies.
Platform literacy education becomes crucial. We need to prepare young people to recognize explicit and implicit, verbal and multimodal tropes and narratives in their actual information environments—their social media feeds, algorithmic recommendations, and comment section exposures. This means developing critical consumption skills for platform-mediated content and recognizing when legitimate political discourse slides into antisemitic territory.
The Path Forward
Antisemitism remains a minority position even among young people—57% rejected antisemitic propositions. The challenge is preventing the 43% from growing. Rosenberg’s call for leadership that strengthens consensus rather than pandering to extremes requires understanding what threatens that consensus: infrastructure and persuasion working in tandem.
We need to address both dimensions. Yes, platforms must be reformed—when they surface conspiracy theories faster than education can counter them, when coded language evades moderation, when algorithms amplify conflict, consensus erodes through structural mechanisms. But we also need direct engagement with young people in primary schools, secondary schools, and universities. Educational interventions need to account for the fact that antisemitism functions as an unusually complex and adaptive belief system—one that undermines what young people typically strive for: understanding the world at its foundations (in its Grundfesten) and clearly distinguishing between good and evil. Antisemitism offers false clarity, simple explanations for complex crises, and the illusion of moral certainty while systematically distorting both understanding and judgment.
The generational gap with no partisan gap only makes sense if antisemitism is being reproduced through shared communication environments while political camps are busy externalizing blame sideways. This is the puzzle Rosenberg’s data present, and it’s the puzzle platform-mediated communication conditions solve.
The Decoding Antisemitism expansion currently underway at NYU, the numerous activities of AddressHate, and many allied initiatives here in the US represent efforts to build infrastructure for structural intervention and policy impact. Rosenberg’s piece calls attention to an urgent problem—our research provides tools for addressing it. The algorithmic escalator is real. Building off-ramps is no longer optional—it is a structural responsibility.