CyberWell is the leading nonprofit organization tracking online antisemitism across multiple languages and platforms worldwide. We work directly with social media companies to reduce Jew-hatred at scale by leveraging advanced technologies and AI to deliver actionable, platform-ready solutions and digital safety insights. CyberWell’s real-time antisemitism compliance service and alert system are designed in alignment with the policies, operations, and content moderation frameworks of major technology companies, enabling effective implementation by industry stakeholders. Our mission is to strengthen the enforcement and improvement of hate speech policies and community standards, making digital spaces safer for Jews and everyone globally.
Since May 2022, CyberWell has continuously monitored antisemitic content on Facebook, Instagram, X, TikTok, and YouTube in both English and Arabic. We serve as a trusted partner for Meta (Facebook and Instagram), TikTok, and YouTube, enabling us to escalate policy-violating content and advise moderation teams. Our flagship initiative is the first ever open database of online antisemitic content, used by academics, policy makers, journalists, and leaders in the Jewish community to empower their efforts with data. This report sums up a portion of our findings for 2025.
| 2025 | 2024 |
|---|---|
| World Domination | World Domination |
| Scapegoating Jews for Global Tragedies | Rothschild |
| Blaming Jews for Committing Attacks on Themselves | Synagogue of Satan |
CyberWell’s monitoring of antisemitic content in 2025 identified three primary narrative categories shaping online antisemitic discourse: classic world domination conspiracies, scapegoating of Jews, and Conspiratorial Self-Victimization. While all three narratives were prevalent during the year, they differ significantly in how they are defined in platform policies and, as a result, how consistently they are enforced or actioned by platforms.
The classic conspiratorial trope of Jewish world domination – claims that Jews secretly control governments, institutions, or global events – remained the most dominant form of antisemitic rhetoric in 2025. This narrative has been central to antisemitic ideology for centuries and continues to circulate widely online. However, removal rates for this category are comparatively high (59%), largely because it is explicitly defined as antisemitic and policy-violating across all major social media platforms.
In contrast, scapegoating narratives and Conspiratorial Self-Victimization are generally not addressed in most platform policies, with TikTok as a notable exception.
Scapegoating narratives attribute responsibility for real-world crises, violence, or disasters to Jews as a collective. CyberWell documented this narrative across a wide range of contexts in 2025, including shootings (the Charlie Kirk assassination), geopolitical events (the U.S. intervention in Venezuela), and even natural disasters (the California wildfires). More broadly, scapegoating rhetoric reliably emerges in the aftermath of nearly every major global event, framing Jews as responsible actors regardless of evidence.
Conspiratorial Self-Victimization (CSV) refers to claims that Jews or Israelis orchestrate antisemitic attacks against themselves in order to manipulate public opinion or gain political or financial advantage. CyberWell observed a sharp increase in CSV following the October 7 attacks and has since identified this narrative in connection with 11 violent incidents, including the attack on Bondi Beach. As highlighted in CyberWell’s 2024 Annual Report, CSV did not remain confined to the immediate post-October 7 period but has continued to resurface after nearly every violent attack targeting Jews or Israelis.
From an enforcement perspective, the lack of clear policy definitions for scapegoating and Conspiratorial Self-Victimization has resulted in substantially lower removal rates. Although scapegoating narratives generally appear to be removed at a rate of 50%, in practice, removal typically occurs only when the content overlaps with established antisemitic tropes already covered by existing frameworks. When scapegoating appears as a standalone narrative, it is generally not removed. Removal rates are even lower for Conspiratorial Self-Victimization – only 37% across platforms – with TikTok as the only platform that consistently actions and removes this content.
Along with scapegoating narratives and CSV, CyberWell continues to observe a prevalence of posts falling within the first example of the IHRA (International Holocaust Remembrance Alliance) working definition, which includes content that glorifies, justifies, or explicitly calls for violence against Jews. This category represents the most overt and dangerous form of online antisemitism and frequently appears in the immediate aftermath of violent attacks against Jews or Israelis. In numerous cases, posts vetted by CyberWell referenced real-world attacks in a celebratory manner or frame violence as justified, desirable, or worthy of repetition – sometimes issuing proactive calls for further violence.
Taken together, these patterns reveal a reinforcing cycle of violence that defined online antisemitic discourse in 2025. Violent attacks against Jews are followed by Conspiratorial Self-Victimization narratives that deny Jewish humanity and reject the recognition of Jewish victimhood. These narratives coexist with - and often give way to - explicit glorification, justification, and incitement of further violence. Rather than relying solely on abstract conspiracies or generalized stereotypes, antisemitic actors are increasingly anchoring their rhetoric to specific real-world events, actively seeking to identify - or fabricate - a “Jewish connection” to contemporary crises, tragedies, or acts of violence, including those in which Jews themselves are the victims. When such event-driven narratives remain online without consistent enforcement, their heightened visibility and virality significantly increase the risk of further harm to the already targeted Jewish community.
CyberWell urges all platforms to acknowledge this shift and explicitly prohibit scapegoating and Conspiratorial Self-Victimization within their digital policies and enforcement frameworks.
Although the average removal rate of antisemitic content across platforms increased modestly - from approximately 50% in 2024 to 52.4% in 2025 - this number masks substantial changes in platform-level enforcement outcomes from 2024 to 2025, detailed below.
TikTok recorded the most dramatic increase, with removal rates rising from 65.1% in 2024 to 88.81% in 2025 (+23.7 percentage points). This marks TikTok as the platform most consistently enforcing its policies in this dataset.1
Meta2 also showed a clear upward trend, increasing from 49.5% in 2024 to 57.31% in 2025 (+10.3 percentage points), crossing the threshold into removing a majority of content reported by CyberWell for the first time since CyberWell began monitoring in 2022 and reflecting more systematic enforcement.3
YouTube’s removal rate nearly doubled, rising from 17.5% in 2024 to 34.17% in 2025 (+16.7 percentage points).4 This increase should be taken in context: 2025 was the first year in which CyberWell operated as a priority flagger for YouTube. CyberWell joined this program in the fourth quarter of 2025 and then retroactively reported all content identified during the year through YouTube’s priority escalation channel. The increase in removal rate likely reflects this new status and reporting avenue, rather than a fundamental change in YouTube’s underlying moderation standards.
By contrast, X experienced a sharp decline in enforcement, with removal rates falling from 54.2% in 2024 to 29.46% in 20255 (-24.7 percentage points), making it the platform with the lowest removal and enforcement rate in the dataset. This decline corresponds with X’s increasing reliance on visibility-limitation measures rather than removal. While framed as a “freedom of speech, not freedom of reach” approach, this strategy has proven insufficient. Despite visibility restrictions being implemented (often only applied after the content gained widespread exposure) the flagged content continued to circulate widely, accumulating almost 30 million views in this dataset alone.
Footnotes:
Between 2024 and 2025, enforcement of content falling under the fourth and fifth examples of the IHRA working definition – Holocaust denial and distortion – diverged sharply across platforms. While several platforms demonstrated improved moderation outcomes, others continued to treat this content as “non-actionable”, resulting in uneven enforcement across the online ecosystem.
In 2025, CyberWell vetted 185 antisemitic posts related to Holocaust denial and distortion. The data indicates a notable shift in the prevalence of this content - after a steady annual 3% increase between 2022 and 2024, the proportion of posts containing Holocaust denial and distortion declined significantly in 2025, dropping from 11% in 2024 to less than 8% in 2025.
** Some posts contain both IHRA examples 4 & 5 and are therefore counted more than once.
This decline appears to be driven by two overlapping dynamics. First, denial narratives have increasingly shifted away from historical events toward contemporary atrocities against Jews. Since the October 7 attack, denial and distortion of the Hamas massacre have remained highly prevalent, even two years later. These narratives include outright denial that the attack occurred, denial of specific elements such as use of sexual violence as a weapon of war, and claims that Israel or Jews orchestrated the events - an expression of Conspiratorial Self-Victimization that mirrors classic Holocaust distortion patterns while targeting recent events.
Second, improved enforcement against explicit “classic” Holocaust denial appears to have incentivized more evasive forms of Holocaust-related hate speech designed to bypass moderation. These include implicit mockery of victims, coded references to “six million,” (the documented number of Jewish Holocaust victims), euphemisms such as referring to Hitler as a “painter”, and AI-generated or stylized content that portrays the Holocaust or Nazi figures in a positive, ironic, or trivializing manner. While such content may not always meet narrow definitions of Holocaust denial or distortion, it nevertheless constitutes Holocaust-related hate speech. Platforms frequently fail to identify the implicit, cynical, or coded language embedded in these formats, encouraging antisemitic actors to adopt them in place of more explicit denial. This issue was exacerbated by the rollout of generative AI platforms that produced content with coded Holocaust hate speech and distortion. Examples include the “Caust”, a Pixar-style movie trailer generated with Sora, which trivializes the Holocaust by mocking victims and turning its atrocities into child-friendly entertainment, as well as similar AI-made trailers targeting children that glorify Hitler.
At the same time, enforcement actions against Holocaust denial and distortion increased across most major platforms in 2025, continuing a trend observed in previous years. The most significant positive change occurred on YouTube, where removal rates rose substantially following CyberWell’s designation as a priority flagger in 2025. For the first time, all identified Holocaust denial and distortion content was escalated through YouTube’s priority reporting channel, resulting in markedly improved enforcement outcomes, from just 30.77% in 2024 to 58.33% in 2025.
** Some posts contain both IHRA examples 4 & 5 and are therefore counted more than once.
The primary exception remains X, where enforcement against Holocaust denial and distortion declined sharply. In 2025, CyberWell observed that X permitted content that questions or casts doubt on well-documented Holocaust facts – including the scale and mechanisms of the genocide – so long as such posts do not explicitly target Jews as a protected group. As a result, significant portions of Holocaust denial and distortion content are not considered policy-violating and therefore are not actioned. Moreover, even when content does meet the platform’s threshold for a violation, X often relies on visibility-limiting measures rather than removal, allowing Holocaust denial and distortion to remain accessible, albeit with reduced amplification.
From both a policy and enforcement perspective, these developments are closely interrelated. Holocaust denial and distortion are among the most clearly and consistently prohibited forms of antisemitism across major platforms. This high level of policy clarity stands in contrast to other current antisemitic narratives that lack explicit policy provisions and frequently fall into enforcement gray areas, likely contributing to more consistent and decisive moderation in this category.
A striking point of comparison to Holocaust denial and distortion is Conspiratorial Self-Victimization and the denial of recent violent antisemitic attacks. While these narratives share core elements of denial and distortion of real-world harm, they differ in a critical respect: they operate within an ongoing cycle of antisemitic violence rather than as retrospective reinterpretations of historical events. Unlike Holocaust denial and distortion – which are explicitly prohibited and therefore more consistently enforced – these contemporary denial narratives remain largely absent from platform policies. This gap is especially consequential because such narratives emerge in real time following violent attacks, reframing those events as hoaxes or false-flag operations and denying Jewish victimization as it unfolds. In this form, these narratives pose a more immediate risk than Holocaust denial and distortion alone, while continuing to represent a significant blind spot in platform enforcement frameworks.
In 2025, CyberWell identified two major evolving mechanisms through which antisemitic content spreads online: AI-generated content and coded language.
According to CyberWell’s analysis, AI-generated content took a growing role in lowering the barrier to creation and enabling the rapid dissemination of antisemitic narratives across platforms. AI tools enable antisemitic actors to quickly generate and repackage content at scale, amplifying existing tropes and adapting them to new events with minimal technical skill.
Much of this AI-enabled antisemitism does not rely on explicit hate speech or overt glorification of violence. Instead, it often uses humor, cynicism, and implicit framing to evade enforcement, embedding antisemitic narratives within memes, short-form videos, animations, and other highly shareable formats. These approaches broaden exposure – often reaching audiences not actively seeking extremist content – while exploiting moderation systems that struggle to identify hate when it is expressed indirectly.
This dynamic is closely linked to the broader phenomenon of coded language, which CyberWell documented as widespread and escalating in 2025. Antisemitic actors increasingly rely on antisemitic “dog whistles”, including linguistic substitutions, numeric shorthand, euphemisms, and symbolic imagery, to circulate antisemitic tropes while bypassing detection. These coded forms are particularly prevalent in youth-oriented or adjacent digital spaces, including gaming content, where antisemitic narratives are embedded within seemingly unrelated content and normalized through repetition.
Taken together, AI-generated content and coded language facilitate the cross-narrative distribution of antisemitic rhetoric and allow it to persist even as platforms improve enforcement against more overt hate speech. They therefore represent a growing challenge for platforms, underscoring the limits of moderation systems that remain primarily calibrated to detect explicit hate rather than evolving, implicitly framed antisemitic narratives.
Establishing clear, updated, and consistently applied policy guidance to address these emerging challenges and antisemitic tropes – through consultation with trusted experts on antisemitism, such as CyberWell – would strengthen enforcement and help prevent harmful content from slipping through moderation gaps.
CyberWell continues to monitor the online space for antisemitic content, identifying trends and reporting directly to our partners at the major social media platforms, all in the efforts of ensuring a safer digital future for the Jewish people and everyone.
© 2026 CyberWell. All Rights Reserved.
