The Changing Face of Antisemitism in the US

This is a relatively long article for Heterodox STEM and the theme is puzzling. What does the evolution of antisemitism have to do with STEM? The answer is that elite universities have become hotbeds of antisemitism due to woke/DEI/Islamist influence, that the discrimination has provided the Trump administration legal grounds for rolling this influence back, and that funding for scientific research is caught in the crossfire. One countercharge is that campus antisemitism is blown way out of proportion relative to white supremacist agitation in the boondocks, in which case universities should likely just dig in their heels. I will show that the countercharge is false by drawing on data from a surprising source: an organization that is proudly supportive of DEI and actively hostile to white supremacy.

Recent terrorist attacks on Jews in the US have been met with a mixture of shock, dismay, and attempts at denial. Their leftist/Islamist face defies the widespread belief that antisemitic extremism is mainly right-wing white nationalist. One influential promoter of that belief is the Anti-Defamation League (ADL). In April 2022 it noted “an acute threat of antisemitic terrorism … which is overwhelmingly from right-wing extremists and in particular white supremacists”. In November 2023, it warned of “a steady rise of right-wing terrorism in the US [since 2008 with] no likelihood of significantly decreasing soon”. Its annual Audit of Antisemitic Incidents, published less than two months ago, refers 13 times to “white supremacist” or “right wing” but never to “left wing” or “Islamist”.

The Audit draws on an ADL-compiled database of Hateful, Extremist, Antisemitic and Terrorist incidents, called “H.E.A.T. Map^TM” (hereafter “HEAT”). The Audit is widely cited in the press and helps mold public consensus. For example, when ChatGPT or Grok are asked for summary data on terrorism, they mainly cite the ADL. Unfortunately, the ADL’s reports are greatly distorted, partly by what they present and more by what they leave out. They grossly exaggerate right-wing transgressions and grossly understate leftist/Islamist transgressions.

Nowadays the most virulent form of antisemitism aims to free Israel of Jews. It is the form observed on elite college campuses, pretending to social justice and draped in keffiyehs. Its impact is greater than all other forms combined and rightist white-nationalist influence there is negligible.

Combining all antisemitic incidents recorded in HEAT for 2024, I estimate 14% Right Wing, 57% Left Wing, 23% Left/Islamist, and 6% Islamist. (The last two categories likely encompass some Black nationalist influence that HEAT obscures). The Right Wing share would drop below 5% if HEAT recorded incidents more fairly and took impact into account. For example, HEAT rated one rightist bomb hoax emailed simultaneously to 29 nearby synagogues as nearly twice as significant as all the hostage posters vandalized in Manhattan over the course of 2024. Also, Right Wing incidents tended to involve small publicity stunts or one-on-one harassment.

The new landscape is also marked by Holocaust inversion. References to Nazi predations now do less to inoculate against Jew-hatred than to add venom to it. Israelis are equated to Nazis. Jews are painted as culprits in their past and future demise

Many readers will find this hard to believe. It seems to defy US history, to ridicule refined political sympathies, to slander the ADL, and to whitewash hateful whites. Let me start by addressing these concerns from a big-picture perspective, without any statistical analysis. While this will not convince anyone that I am right, hopefully it will persuade that I might not be wrong.

Historically, antisemitism took two main forms in the US: Jews as a despicable non-European race and Jews as despicable non-Christians. Both were openly espoused by the far-right Ku Klux Klan. Accordingly, the ADL focused mainly on combating right-wing, nominally Christian, white supremacists. It deservedly won much praise for these efforts.

However, antisemitism was never exclusively a rightist, Christian, or European affair. Once Nazi Germany was defeated, the leftist, atheistic Soviet Union became the main European vector of antisemitism. After initial defense of the new state of Israel, it branded Israel as an outpost of Western imperialism and advertised that worldwide. Israel’s Arab neighbors became cauldrons of antisemitism that mixed Muslim, Nazi and Soviet strains. The toxic brew re-entered the West through Muslim immigrants stirred up by radical Islamists. Thus antisemitism resurged mainly in three forms relatively new to the US: Jews as a despicable European race. Jews as despicable non-Muslims, and Israel as a despicable country.

While most of the Western left initially opposed this obscurantist nonsense, that changed as Marxist “industrial workers as vanguard” gave way to Marcusian “students and professors as vanguard”, who in turn looked for followers among the laggards in global progress. Muslims from the Middle East particularly lagged due to a combination of poor education, high birth rates, discrimination against women, and reliance on oil revenues over human capital development. Over the next few decades, leftist parties came to appease the obscurantism and then to endorse it, especially where voting blocs of Muslim immigrants grew. The ostensible justification was that oppressed peoples lacked agency and indeed were goaded into backwardness by Israel.

These trends were long visible in France, Canada and the UK, which host the most Jews outside Israel and the US. US leftists were generally slower to make this shift and more restrained, as Soviet/Muslim influences were weaker and pro-Jewish influences stronger. However, the October 7 massacres threw down a gauntlet that many US leftists felt obliged to publicly pick up, no matter how many truths they had to trample.

The ADL is not blind to this. It disapproves of the newer forms too. Yet the ADL rarely combats them with the zeal that it combats the KKK-type forms. It chases mice out of the big rooms of antisemitism while tiptoeing around the elephants.

Why the double standard? One factor was inertia. The ADL kept focusing on Nazi/KKK-type threats long after they faded. Another was quest for approval from other social justice warriors, who considered themselves morally superior to others. (Never mind the difficulties of discerning others’ hearts or the evidence that the road to hell is paved with good intentions.) This led the ADL to treat antisemitism on the left far more tenderly than antisemitism on the right and to try to appease the appeasers.

For proof that this has not worked, I refer to Harvard’s 311-page report on internal “Antisemitism and Anti-Israeli Bias”. The many incidents it described, mounting over many years, rarely if ever made it to HEAT despite Harvard’s huge influence on American society. Below is a picture from an assigned reading for a course at Harvard’s Graduate School of Education, intended to train future leaders of school systems across the US. It depicts the ADL as the highest level of coded white supremacy, on which the apex of overt hate crimes and genocide rests. It is an old trope: the Jews as master manipulators even of the causes they claim to support.

While the ADL’s double standard tilts left, the center-left ultimately stands to lose most from the distortions. The US center-right lost ground in the 1960s partly because it was too prone to blame civil-rights agitation on communist infiltration. Israel’s center-left lost ground in the 2000s partly because it was too prone to blame the government for Palestinian rejectionism and intifadas. In the present-day US, tarring opposition to DEI as white supremacist or lauding Islamists as freedom fighters fuels a rightward reaction.

As for my statistical estimates of rightist or leftist/Islamist influence, I did not arrive at them lightly. I applied numerous filters, reviewed every characterization for possible misinterpretation, and tried hard to be even-handed. Granted, readers will naturally be skeptical of my assurances; I worried too since I might have fooled myself. To obtain more neutral perspectives, I fed HEAT data to two different AI agents without sharing my own evaluations. For transparency I will present both the various results and the methodologies that generated them.

Background on HEAT

Every record in HEAT includes date, city, state, type of incident, and brief description. An additional field, with value inferred from the description, indicates whether the incident was Israel/Zionism-related. HEAT recorded 9,582 incidents for 2024, the highest ever, up slightly from 2023 and more than 2016-2020 combined.

Given the ADL’s history and constituency, one would expect HEAT to focus mainly on antisemitism. However, the focus is so narrow that the name HEAT is a misnomer. In 2024, antisemitic incidents comprised 97.6% of the total. “White Supremacist Events” comprised 90% of the rest. The remaining 42 incidents involved terrorist plots and extremist-related killings.

HEAT contains additional fields for perpetrating group and ideology but most of their entries are blank. Where the group entry is filled, it nearly always refers to right-wing white nationalists; the Black Hebrew Israelites with 3 incidents are the sole exception. Only 12 incidents mentioned an ideology that is not White Supremacist, with 8 characterized as Islamist, 3 as Left Wing, and 1 as Other.

The 2024 Audit defines antisemitic incidents “as vandalism of property, or as harassment or assault on individuals or groups, where either 1) circumstances indicate anti-Jewish animus on the part of the perpetrator, or 2) a reasonable person or group of people could plausibly conclude that they have been victimized due to their Jewish identity”. The Audit explicitly excludes “legitimate political protest, support for Palestinian rights, or expressions of opposition to Israeli policies”, “burning or desecrating Israeli flags”, and some “physical scuffles or verbal insults between pro- and anti-Israel protestors”. This leaves a big grey area requiring judgment calls. For example, ADL generally views BDS resolutions as antisemitic but does not include them in HEAT. For the rest of this paper, an “antisemitic incident” means whatever HEAT labelled as such.

The limited information on perpetrators and the thin coverage of extremism are two big shortcomings of HEAT. Two other big shortcomings are the inattention to mainstream news or online commentary and the disregard of “reach”, i.e., how many likely victims with what impact. None of these shortcomings are fully avoidable. Many perpetrators cannot be identified. The ADL lacks the bandwidth to cover all extremism in media. “Reach” requires subjective judgments with big margins of error. However, ADL could easily cover much more than it does with the help of AI and present it far more fairly.

Exaggeration of Right-Wing Extremism

HEAT’s shortcomings pose the following potential danger. Imagine a small group of unsophisticated extremists, with little funding or protection from domestic or foreign governments, little traction among the intelligentsia, and little mass appeal. Short on numbers and substance, it recruits at various small venues, posts a few banners on overpasses, and distributes a few fliers, relying on half a dozen reasonable phrases and a few shockers. If sufficiently offensive to ADL trackers, it might gain numerous mentions in HEAT despite minimal reach. We might even see some symbiosis, where each side gains through outspoken condemnation of the other.

The danger is not just hypothetical. HEAT gives tiny right-wing white-nationalist groups far more prominence than they deserve. They were held responsible for 13% of all HEAT incidents in 2024. The biggest is the Patriot Front with an estimated 200-250 members, little funding, no significant support in universities, no known government backing, and no known para-military affiliation. Of its two charges over the past six years for physical assault, one in 2022 was recently slapped with a $2.7 million judgment. HEAT gave the Patriot Front sole discredit for 799 incidents (over three per member) in 2024 and shared discredit for 22.

For comparison, consider Students for Justice in Palestine (SJP). By the ADL’s own account, the SJP has more campus chapters than the Patriot Front has members. It draws on abundant financial support and training from both leftist and Islamist organizations. It is linked to terrorist organizations. It has significant traction in universities. It frequently calls for the annihilation of Israel. It frequently excuses or defends the October 7 massacres. It helps organize the intimidation of Jewish students on their campuses. Yet the SJP garnered fewer mentions in HEAT for 2024 than the Patriot Front. And nowhere does HEAT label the SJP an extremist organization.

Or consider the Council on American Islamic Relations (CAIR), with over 60 paid staff, 300 active volunteers and an annual budget of over $1 million. The ADL faults its leaders for terrorist support, comparing Israel to Nazi Germany, accusing Israel of genocide, calling for Israel’s destruction, and indulging in antisemitic tropes. HEAT held CAIR partly responsible for 39 incidents in 2024, less than 5% of the Patriot Front number.

Or consider the Nation of Islam, which the ADL describes as “the largest Black nationalist organization in the US, [which] has maintained a consistent record of antisemitism and bigotry since its founding in the 1930s”. Despite its estimated 35,000 members and a weekly publication that lambasts Israel about as harshly as CAIR does, HEAT mentioned the Nation of Islam in only 2 incidents for 2024.

Besides the Patriot Front, the ADL gives special attention to the rhetorically flamboyant Goyim Defense League. The Goyim Defense League is much smaller than the Patriot Front and no better connected to external support or violence. Its main media platform has less than 15,000 followers. Yet HEAT flagged it for 220 incidents in 2024.

Judging from media reports, extremist white-nationalist groups appear to have less than a thousand members in total with less than 100 thousand followers. From a historical perspective, this marks tremendous progress. In 1925, over 3% of the US population was estimated to belong to the Ku Klux Klan, a rate which corresponds to over 10 million people today. While no country tolerated formal racial subordination longer, arguably no country has worked more diligently to stamp it out. The overwhelming majority of Americans no more want to strip Blacks of equal rights than to reimpose slavery. The core disputes involve preferential rights for Blacks, which supporters view as just and useful reparations and critics view as unfair and counterproductive.

This is evident in the behavior of the white-nationalist organizations themselves. While HEAT labels nearly all of them White Supremacist, none of them affirm that in the incidents described. The label appears only pejoratively, either to condemn white nationalism or to identify Zionism with it. And the labelling clearly hurts. Only a few dozen events, all of them tiny, involved groups with neo-Nazi or KKK-type names. White-nationalist demonstrators are frequently outnumbered by white protestors, which would have been unimaginable in the 1920s.

Critics try to refute this progress in two main ways. The first claims “dog whistles”, where white supremacy is so implicit in white-nationalist appeals that it need not be expressed outright. However, 99% of Americans evidently reject the dog whistles too or membership would be much higher. The second claims “white fragility”, where rejection testifies to racist denial of inner racism. This is not a testable hypothesis; it proves by assertion. But if true, it implicitly agrees that openly white-nationalist organizations are overrated in importance.

The ADL’s exaggeration of white-nationalist extremism is evident in the HEAT incidents themselves. Let’s take the treatment of SJP and the Nation of Islam as benchmarks. The ADL insists that no SJP or Nation of Islam chapter should necessarily be tarred by the affiliations of national SJP or National of Islam leadership, by the statements or actions of other chapters, by what their websites or charters indicated, or even what the chapter itself did last week. None of their events get recorded in HEAT simply for the fact of SJP or Nation of Islam participation.

In contrast, as previously noted, HEAT has a category called White Supremacist Event. In over 100 of these events, none of the described activities or slogans were prima facie subversive. There is nothing inherently hateful about “meetup” or “meetup and training” (42 events with no other offense described). There is nothing inherently white supremacist about hiking or camping out (15 events), visiting an apolitical festival or war memorial (11 events), or attending a sparring competition (7 events). There is nothing inherently extremist about banners for “Strong Families, Strong Nations” (8 mentions), “America First” (8 mentions), protection of borders (2 mentions), or “Help Appalachia! End Foreign Aid”. “Reclaim America” (17 mentions) sounds edgier but can be interpreted in various ways. The offenders were cited for who they allegedly are, not for what they directly said or did.

White nationalists were also cited in a few dozen incidents for protesting mistreatment of whites, with no attendant calls for violence or revenge. No protests by Black nationalists of mistreatment of Blacks would be recorded in HEAT even if far more inflammatory. The obvious counterargument is that white and Black positions in US society are not analogous. White racism clearly deserves extra attention and repudiation. However, it is both categorically wrong and politically dangerous to brand white discontent with reverse discrimination as extremist. When peaceful complaints about real policies are dismissed, it inevitably boosts sympathies for true extremists, who allege a vendetta against whites and sneer at moderation.

In fairness to the ADL, it likely intends to warn of right-wing white nationalists more generally. However, the evidence they offer is thin. Their events cited in 2024 averaged about a dozen members each. Of their reported antisemitic vandalism, 98% involved the placement of stickers, most of which were likely removed before more than a few dozen people viewed them. Their antisemitic harassments occurred disproportionately in small rural towns that few Jews frequent, with no reported physical assaults. The total number of antisemitic incidents HEAT attributed to them in 2024 shrank by 17% from 2023.

Yet the reach of right-wing antisemitism in the US expanded greatly in 2024. Tucker Carlson’s interview with Darryl Cooper, condemned by 24 Democrat congressmen as a “Nazi apologist and Holocaust denier”, attracted 33 million views. Candace Owens, another right-wing pundit with 4.5 million online subscribers, promotes more antisemitic tropes than Carlson and is more openly hostile to Israel. No white-nationalist label fits as Owens is Black. And since HEAT ignores online postings, neither Carlson nor Owens appear in it.

In short, the ADL’s obsession with KKK-type relics is woefully misplaced. The US is quarter way through a new century with new challenges played out in new forums with new enemies. New perspectives are needed.

Terrorist Plots, Extremist Murders, and Police Shootouts

The ADL’s prediction that most terrorism in 2024 would be right wing did not pan out. HEAT listed 21 terrorist plots in 2024. Of the 7 that targeted Jews or support for Israel, 4 were Islamist, 2 were Left Wing, and 1 was Right Wing. Of the two attempted assassinations of Trump in 2024, HEAT ignored Crooks’ attempt and classified Routh’s attempt as Other (with Routh described as a former Trump supporter whose motivations were unclear). From my perspective:

• Both assassination attempts against Trump should be included and classified as Left Wing.

• The assassination of United Healthcare CEO Brian Thompson, for which Luigi Mangione is being tried, should be included and classified as Left Wing.

• A plot by Harun Abdul-Malik Yener, an aspiring ISIS-friendly martyr, should be classified as Islamist. (ADL calls him Right Wing on the grounds that he applied to the Proud Boys but the latter rejected him as an ISIS-friendly martyr.)

• Four foreign terrorist plots linked to ISIS or Iranian operatives, stymied by Federal arrests in the US, should be included and classified as Islamist.

With those corrections, the breakdown of terrorist plots was 12 Islamist, 9 Right Wing, and 6 Left Wing. The DHS Homeland Threat Assessment also mentions numerous other terrorist attacks that were foiled but, as often occurs for security reasons, does not report details. Nor does HEAT include mass school shootings or planned shootings. In short, white supremacist terrorism no longer appears to merit an overriding concern.

At first glance, the 21 extremist murders or shootouts with police listed by HEAT for 2024 bring right-wingers back to the fore, as they were accused of all of them. Here is a representative summary from USA Today:

“For the third year in a row, all of the extremism-related murders last year were committed by far-right extremists, the ADL researchers found. The murders in 2024 included eight killings involving white supremacists and five deaths at the hands of far-right anti-government extremists. At least two of those 13 murders were committed last year by members of the so-called ‘Sovereign Citizen’ movement, a collection of thousands, if not tens of thousands, of anti-government radicals who believe they're not subject to local or national laws or authority.”

Similar reports were widely circulated. Both ChatGPT and Grok affirmed that no Black, Islamist, or Leftist extremist murders or shootouts occurred in 2024 and cited the ADL for evidence.

However, these summaries are highly misleading. Of the 8 sovereign citizens involved in police shootouts, 5 were Black. At least 2 were adherents of the Moorish Temple, a Black-only religious group akin to the Nation of Islam. Of the 13 people killed by extremists in the 21 incidents, 2 were Black (one a Black police officer slain by a Black sovereign citizen) and 6 were white family members or acquaintances.

Furthermore, nearly all the incidents involved prison breakouts, family disputes, or psychotic episodes, e.g., Justin Mohn decapitating his father and displaying the severed head on video while he ranted on mostly right-wing themes. If these are included, then in fairness the ADL should ascribe every killing by convict or crazy to the ideology espoused. A glaring example from 2023 was the mass shooting at a private Christian school in Nashville. Transgender alumn Audrey Hale left a manifesto testifying to anti-rightist inspiration for killing 6 people. Yet HEAT doesn’t reference it, let alone call it terrorism or label it Left Wing (which it should, given that it classified a thwarted bombing in 2024 by a self-described non-binary “neurodivergent goth” as Left Wing).

Antisemitism Focused on Israel

2024 marks the first year that a majority (58%) of antisemitic incidents in HEAT refer to Zionism or Israel. However, Zionism has come to have three meanings. The original meaning was that Jews should re-congregate in what is now Israel. The second and most common meaning today is defense of Israel’s existence as a mostly Jewish country. A third meaning pejoratively identifies Jews as Zionists while nominally denying that the label is antisemitic. Here the Israel connection seems peripheral to classical hatred of Jews.

Consider the slogan “Zionists out of government”, a favorite of the Patriot Front. Since the government in question is the US and the great bulk of its operations have little to do with Israel, it begs the translation “No Jews or Jewish sympathizers in government”. While the sloganeers hate Israel too, the slogan itself doesn’t rule out peaceful coexistence. Conversely, the slogan “Globalize the Intifada” is patently hostile to Israel without mentioning Israel nor Zionism.

After close examination, I redivided HEAT’s 9,354 non-terrorist antisemitic incidents in 2024 according to whether they are directly related to Israel. This section discusses the Israel-related antisemitic (“IRAS”) incidents, which slightly outnumbered the rest. Here are the main calumnies:

• Israelis are illegal settler-colonists, said by people endorsing open borders in the West.

• Israelis should go back to Europe, although a huge share come from the Middle East.

• Israel is genocidal in Gaza, despite Gaza’s sustained rapid growth in population.

• Israel practices apartheid, when minorities have far more rights than in neighboring countries and latter don’t even feign to welcome Jews.

• Hamas fights Israel for freedom and justice, when its record even in Gaza alone is savage and unfailingly repressive.

While Jew-hatred finds support across the political landscape, these accusations against Israel appeal far more to left wingers than right wingers, far more to Black nationalists than to white nationalists, and far more to Islamists than to radicals of other faiths. In the US, the white right is much more concerned about immigration, racial preferences, and sexual teachings in schools than about Israel’s existence. A growing share views Israel as an important US ally and some conservative Christians are more Zionist than liberal Jews. The term “white” itself, once identified only with northern Europeans, has evolved in common parlance (far right excepted) to include southern Europeans and Ashkenazi Jews.

In contrast, Israel’s allegedly corrupt existence provides a central rallying cry for both the far left and Islamists. It combines colonialism, capitalism, Western finance, humiliation of Arabs, and insult to Islam. Israel also looks enticingly vulnerable thanks to millions of resentful Palestinians on its borders, well-funded armed resistance, and wavering Western support.

Hence, we would expect IRAS incidents to be far more leftist or Islamist in ideology than rightist. HEAT data on location supports that. I asked ChatGPT to rate the leading cities (including boroughs of NYC, which HEAT distinguished) by political orientation, using its choice of criteria. It decided to use presidential vote margins in 2020 and 2024, voter registration data, local election results, and policy positions like sanctuary status or policing reforms. Here are headline results:

• Over 10% of IRAS incidents occurred in Manhattan or Brooklyn, with both rated Far Left.

• 33% of IRAS incidents occurred in the top 14 cities or NYC boroughs, all rated Left or Far Left (in order: Manhattan, Los Angeles, Chicago, Philadelphia, Washington, Brooklyn, San Francisco, Ann Arbor, Seattle, Atlanta, Boston, Denver, Portland, San Diego).

• 50% of IRAS incidents occurred in the top 38 cities or NYC boroughs, with none of them further right than Center-Left.

Local governments in these locations tend to be proudly anti-rightist. They abhor the hostility to immigrants, Blacks or gays that right-wing rallies against Israel would almost surely invoke too. Right-wing demonstrators there risk big counter-protests, widespread ridicule, and occasional assault. From a right-wing white-nationalist perspective, these are enemy territories not worth messing with apart from occasional guerrilla stunts intended to annoy.

31% of IRAS incidents occurred at colleges and universities, which usually lean farther left than local governments, host many students from countries hostile to Israel, and combat transgressions against DEI vigorously unless the victims are Jews. The cities mentioned above owed their high IRAS rankings in part to college activists. But many campuses in conservative areas were havens of anti-Israel protests too. In Alabama, as many leftist IRAS incidents occurred on college campuses as white-nationalist IRAS incidents elsewhere.

ChatGPT concurred. It found “no widely documented or prominent openly right-wing rallies against Israel” in 2024 either on US college campuses or in US cities. It attributed that to

• the overwhelmingly pro-Israel stance of the mainstream US right, driven by “evangelical Christian voters, Republican foreign policy, and opposition to progressive causes”.

• far-right groups’ “lack [of] organizational support to stage rallies in politically hostile, Left-leaning cities”.

• public anti-Israel protests “saturated with left-wing organizers leaving little space for right-wing groups to co-opt or initiate similar actions”.

• any right-wing presence “more likely to manifest as counter-protests supporting Israel”.

HEAT downplays this. While 99% of IRAS incidents leave the fields for perpetrating groups and ideology blank, many could be filled in from the descriptions. They mention the SJP 781 times, Party for Socialism and Liberation (PSL) 754 times, Jewish Voice for Peace (JVP) 314 times, Palestinian Youth Movement 304 times, Democratic Socialists of America 169 times, Students for a Democratic Society 147 times, ANSWER Coalition 127 times, American Muslims for Palestine 109 times, Freedom Road Socialist Organization 95 times, Palestinian Community Network 92 times, Within Our Lifetime 67 times, Witness for Peace 53 times, Al-Awda 49 times, Healthcare Workers for Palestine 43 times, and Council on American-Islamic Relations 39 times, Palestinian Solidarity Committee 35 times, and groups named Communist or Revolutionary 28 times.

Leftist or Islamist groups organized nearly 2400 IRAS incidents in 2024, or just over half of the IRAS total. That number exceeds the total number of antisemitic incidents reported by ADL for any year before 2021. Yet the ADL, which is fastidious about right-wing white-nationalist activities, has never provided a ready tabulation and glosses over the size.

After combing repeatedly through the remaining IRAS incidents. I found only about 60 that struck me as right-wing white nationalist. They criticized Jews as pro-immigration, anti-white, anti-Christian, or anti-US. Most reference to the Middle East was tangential. For example, bomb threats were emailed to 29 synagogues that claimed solidarity with Palestine but explained “Heil Hitler! I am sick and tired of you deceitful, lying Jews destroying this country through mass immigration and degeneracy!”

While hundreds more messages had no explicit ideological stance, their locations and broader context strongly favor leftist or Islamist interpretations. Rightists would need to add something distinctive to stand out from the crowd. For similar reasons, I classified only 7 incidents as Black nationalist as no others manifested clear hostility to both Jews and whites.

In total, I classified 2.5% of IRAS incidents as Right Wing, 51% as Left Wing, and 45% as Left/Islamist mix. The rightist share may sound preposterously low given the plethora of references to Nazism or the Holocaust. These were once tell-tale signs of far-right white nationalism. However, in Islamist and Black-nationalist circles Hitler is much admired or excused. Meanwhile, abhorrence of Nazism has morphed on the left into portrayals of Israel as the most Nazi-like regime on earth.

The combination has turned the once-derided swastika into a unifying symbol of anti-Israel wrath. Out of 122 IRAS incidents mentioning swastikas, 70 labeled Israel as Nazi-like. In another 15 incidents, swastikas were used to honor Palestinian resistance to Israel. Where swastikas were used to insult Israel without other explanation, the location was usually a city rated Far Left by ChatGPT. Only 8 IRAS incidents with swastikas had a clearly far-right white-supremacist slant.

Even 2.5% greatly overstates the IRAS reach of far-right whites. Leftist and/or Islamist incidents tended to involve far more people and affect far more Jews. For example, the far-right bomb threat emailed to 29 synagogues was classified as 29 separate incidents even though all the synagogues were in the same county (Westchester NY), the replications were likely noticed quickly, and the very form of delivery suggested a single hoax. In contrast, the vandalizations in Manhattan of hundreds of hostage-related fliers and posters were recorded in only 16 incidents. Nor was any single rally involving multiple offenses cited in two HEAT incidents or more.

AI Perspectives

For a first AI opinion, I asked GPT-4o, the main current ChatGPT model, to infer the ideology that most likely motivated the IRAS incidents, using the categories Right Wing, Left Wing, Islamist, Left/Islamist mix, and Black nationalist. GPT-4o initially responded that “publicly described ADL patterns (2023-24)” led it to expect roughly 40% Right Wing, 25% Left Wing, 15% Islamist, 10% Left/Islamist, and 5-7% Black nationalist. However, analysis of the data quickly changed its mind. Of the 68% of incidents it was willing to classify based on the descriptions alone, leftist and/or Islamist outnumbered rightist ten to one, while classification as Black nationalist was very rare.

I next asked GPT-4o to fold in the information on protests and political leanings that it had provided in other queries. It proceeded to classify all remaining incidents. Interestingly, this did not significantly change the relative shares. Right Wing comprised 9% of the total. However, roughly a quarter of allegedly rightist incidents were leftist or Islamist-sponsored rallies. Another quarter accused Israel of acting like Nazis, with a blended swastika/Star of David as a favored meme. Some distinguished between Judaism and Zionism and urged Jews to be anti-Zionist. Some called for Islamist rule. When questioned, GPT-4o conceded that terms like “Nazi”, “Hitler”, or “swastika” had mistakenly triggered a Right Wing label. It explained:

In your dataset, many such references were not endorsements of Nazism, but comparisons of Israel to Nazi Germany, which is common rhetoric in far-left or left/Islamist discourse [and] not a hallmark of actual right-wing extremism.

This fascinates both for the initial mistake and the ready acknowledgment. Humans make similar mistakes but their acknowledgments are slower and more begrudging. The challenges point to the need for better education about the metamorphosis in Nazi-related rhetoric.

GPT-4o’s corrections raised the Left Wing share to 72% and lowered the Right Wing share to 7%. However, over half the rightist inferences still struck me as badly mistaken. GPT-40 seemed to apply some pecking order of keywords that I did not understand. It also frequently mistook sarcasm for neutral observations or praise.

For a second AI opinion, I presented the same core queries to GPT o3, which uses more advanced reasoning than GPT-4o. Some of its initial classifications came close to mine. Right Wing comprised 2%, slightly less than my 2.5%, while Left Wing comprised 54% versus my 51%. However, nearly all of what I called Left/Islamist it either classified as Islamist (13% total share) or considered too ambiguous to classify. I next asked GPT o3 to compare its classifications to those of GPT-4o, which I identified only as another LLM:

Where do you seem to disagree on logic? Where do you seem to agree on logic but find the other LLM inconsistent in application? Are there any areas where on reflection the other LLM's logic or implementation seems better than yours and you would like to recategorize?

GPT o3 responded at length. It criticized the other LLM mainly for excessively rigid distinctions with combinations handled inconsistently. The main aspects it praised were the “broader Islamist lexicon”, allowance for “hybrid rallies with both Marxist slogans and Hamas chants”, and attention to rally sponsorship by “dozens of minor socialist groups”.

Recategorization kept all Right Wing labels but shifted about a quarter of the rest in a more Islamist direction, resulting in 39% Left Wing, 34% Left/Islamist, and 25% Islamist. Hundreds of the incidents relabeled left/Islamist involved adamantly non-Islamist leftist groups like PSL or JVP denouncing Israel or praising Palestinian resistance without any endorsement of clearly Islamist goals or co-sponsorship with Islamists. Afte noting that to GPT o3, I gave it a choice:

You might recategorize as Leftist for that reason or you might retain the mixed label to emphasize their common current cause. Which do you prefer and why?

GPT o3 opted to reserve the term Islamist for explicitly Islamist aims and to avoid “mask[ing] how large the non-religious, anti-Zionist left actually is”. This led it to recategorize the vast majority of left/Islamist incidents as purely leftist. The revised shares were 2% Right Wing, 68% Left Wing, 5% Left/Islamist, and 25% Islamist. While I noticed some obvious errors (“Gaza's blood is on your hands … Hitler would be proud" is not rightist praise), their share seemed small.

My AI queries boost confidence that the main drivers of IRAS incidents were less than 3% rightist, roughly 25% Islamist, over 50% leftist, and remainder a leftist/Islamist mix. When we take relative impacts into account, rightist influence shrinks to less than 1%. Clear Black nationalist influence seems negligible.

Other Antisemitism

Let us now consider the remaining 49% incidents of antisemitism recorded in HEAT for 2024. While they might have been triggered by hatred of Israel, their descriptions do not reveal it. I will label them Other Antisemitism or OTAS. They differed from IRAS in three striking ways:

• Less than 1% involved public rallies versus over 50% for IRAS.

• 22% were attributed to right-wing white-nationalist organizations versus 1% for IRAS.

• Most other descriptions revealed nothing about the motivating ideology unless combined with other information.

The dearth of rallies indicates that OTAS incidents generally had far less impact than IRAS. The higher white-rightist share accords better with 20^th century history. The thin descriptions beg for better identification of context and sponsors. For example, attackers’ race (which the FBI tries to collect in hate crimes) would help rule out either white nationalism or Black nationalism. Keffiyehs would encourage reclassification as IRAS with leftist or Islamist motivation.

Consider swastikas again. A quarter of OTAS incidents reported their display, usually with no direct information on intent. Observers are forced to draw inferences from better-documented incidents, which nowadays associate swastikas more with left/Islamist stances than with rightist stances. That holds even when Israel isn’t nominally involved. In 2024, groups tagged as white supremacist displayed swastikas in less than 3% of their OTAS incidents whereas swastika frequency in other OTAS incidents exceeded 30%.

As noted earlier, the ADL attributed 22% of OTAS incidents to right-wing white-nationalist groups. I labelled all these as Right Wing, instructed GPT o3 to treat them as correct, and left the remaining ideological labels blank. I then merged the IRAS and OTAS records, added incident dates, submitted the merged file to GPT o3, and prompted it:

Look for similarities in expressions, symbols, location and timing. Do you think these similarities might provide useful clues about the motivating ideologies?

GPT o3 affirmed. It explained that “rare wording or stencil style” often signals the “same author or peer group”, “similar location and tight time-window” often signals a “single wave by one group”, a “campus or city ‘ecosystem’” often indicates “re-use of organizers and hash tags”, and “calendar clustering” often indicates “ideology-specific holidays”. GPT o3 proceeded to classify 60% of previously unrated incidents as Left Wing, 10% as Islamist, and 3% as Right Wing.

Of the roughly 1000 Unclassified incidents that remained, some were clearly anti-Black while others occurred at locations that leaned very leftist or pro-Islamist. GPT o3 defended its caution on grounds that the connections might be accidental but offered to give more weight to circumstantial evidence. The extra weight let it classify nearly all the remaining observations. The final OTAS shares were 26% Right Wing, 63% Left Wing, 10% Islamist, and 1% Left/Islamist.

These inferences invite two objections. The first is that many messages are too cryptic to interpret. If so, then logically the ADL should revert to its Obama-years policy of not classifying them as antisemitic unless they specifically targeted Jews or Jewish institutions. (The 2010 Audit explained that “the Nazi swastika is no longer exclusively used as a hate symbol against Jews; rather, it is used in vandalism incidents targeting others or for its shock value”). But I think inclusion is justified. The whole point of cryptic messages is to make a murky impression. Chat 3o’s categorization focused on how people likely interpreted the messages, which matters more than what the senders meant to convey.

The second objection is that more data is needed on context. That is true but largely the ADL’s fault. For example, HEAT obscures the many Black-on-Hasidic attacks in Crown Heights. When more detail is collected on known assailants or vandalism, inference accuracy will soar.

As noted earlier, the combined estimates for all antisemitic incidents recorded in HEAT for 2024 are 14% Right Wing, 57% Left Wing, 23% Left/Islamist, and 6% Islamist. The last two categories likely encompass some Black nationalist influence that HEAT obscures. The Right Wing share would drop below 5% if HEAT recorded incidents more fairly and took impact into account.

Conclusion

As noted earlier, the combined estimates for all antisemitic incidents recorded in HEAT for 2024 are 14% Right Wing, 57% Left Wing, 23% Left/Islamist, and 6% Islamist. The last two categories likely encompass some Black nationalist influence that HEAT obscures. The Right Wing share would drop below 5% if HEAT recorded incidents more fairly and took impact into account.