What Is the Zionist Occupation Government (ZOG)? Reading the Definitions, the Measured Corpora and the Public Record

ZOG is an acronym for Zionist Occupation Government, also written Zionist Occupied Government, and the Anti-Defamation League, a monitoring organization with an advocacy mission, defines it in its hate-symbol glossary as a white supremacist term for the belief that the United States government is controlled by Jews; the American Jewish Committee dates its first appearance in the United States to the 1970s. Three letters in a comment thread carry no footnote, which is why the question gets asked at all: what does the abbreviation assert, and who has measured it? The claim it compresses has been measured, and the measurement runs in two directions at once: Jikeli and colleagues (2022) found that about 62% of the antisemitic tweets containing “Jews” in their expert-annotated corpus fit the working definition’s paragraph on allegations of Jewish collective power, while the Yale Youth Poll (2025) found two-thirds of US registered voters rejected all three antisemitic statements it tested, including the statement that Jews have too much power.

Key Findings

• The American Jewish Committee dates the term’s first US appearance to the 1970s and names movement sites including Aryan Nations and the National Alliance as places it appeared; the Anti-Defamation League’s glossary defines it as a white supremacist acronym for the belief that the US government is controlled by Jews. Both are definitions published by monitoring organizations with advocacy missions, not measured shares.
• The Yale Youth Poll (2025) found 10% of US registered voters ages 18 to 34 agreed with all three antisemitic statements it tested, against 2% of voters 65 and older, while two-thirds of voters rejected all three.
• Zannettou and colleagues (2020) counted the word “jew” in 1,993,432 of 67,416,903 4chan /pol/ posts collected between July 2016 and January 2018, a share of 3.0%, with the term rising 16.44-fold between the start and end of the dataset.
• Mendelsohn and colleagues (2023) assembled a glossary of over 300 coded terms and found that harmful content containing them evaded automated toxicity detection, direct evidence that coded language is built to pass keyword and classifier moderation.
• The Congressional Research Service reports that under the 2016 Memorandum of Understanding the United States pledged $38 billion over FY2019 to FY2028, subject to congressional appropriation; the House passed the Iron Dome Supplemental Appropriations Act, 2022, by 420 to 9 on 23 September 2021, per the recorded vote at congress.gov.

What does ZOG stand for and what does it claim?

ZOG stands for Zionist Occupation Government, also rendered Zionist Occupied Government, and the Anti-Defamation League’s hate-symbol glossary defines the acronym as a white supremacist term reflecting the belief that the United States government is controlled by Jews. The American Jewish Committee, in its Translate Hate glossary, dates the term’s first US appearance to the 1970s.

The American Jewish Committee’s entry names the movement sites on which the term appeared, including Aryan Nations and the National Alliance. The World Jewish Congress lists it among the conspiracy myths it catalogues, describing the same core assertion: that Jews or Zionists secretly hold the levers of a national government.

Two properties of these definitions matter for reading anything else about the term. First, they are definitional rather than quantitative. Each is a glossary entry published by a monitoring organization with an advocacy mission, recording what the acronym means and how the publishing body classifies it; none is a measured share of any corpus, and none carries a sample, a date range or a confidence bound.

Second, what the acronym asserts is narrower and stronger than criticism of a government’s decisions. The claim is one of hidden control: that the visible institutions of a state are a front, and that the actors setting policy are a group acting covertly on its own behalf. That structure, an occupied state and a concealed occupier, is what distinguishes it from a dispute over a policy, and it is what makes the term legible to a small in-group while reading as an obscure abbreviation to everyone else. The same structural feature places it inside the family of allegations of collective Jewish power that the working definition of antisemitism enumerates, and inside the class of terms that keyword moderation was not built to catch.

Is ZOG confined to the neo-Nazi fringe?

Yes on the acronym, and the sourcing says so plainly: the Anti-Defamation League classifies ZOG as a white supremacist term and the American Jewish Committee locates its early US circulation on movement sites. The belief it encodes is measured on a different scale. The Yale Youth Poll (2025) found 10% of US registered voters ages 18 to 34 agreed with all three antisemitic statements it tested, including that Jews have too much power, against 2% of voters 65 and older.

Those figures are shares of the poll’s respondents who expressed overt agreement, which is the narrowest available measure: two-thirds of voters in the same poll rejected all three statements. The poll also recorded that younger voters were more likely both to hold these attitudes and to hold anti-Israel views.

The distribution across the political spectrum has been measured separately. Hersh and Royden (2022), running several experiments inside an original survey of about 3,500 US adults with an oversample of young adults, found overt antisemitic attitudes rare on the political left and more common on the right, with the highest concentration among young adults on the far right. That finding cuts against the horseshoe framing the authors set out to test.

Why the belief travels beyond the movement that coined the slogan has its own measured literature. Hadar and colleagues (2026), in a model supported across six national samples covering the United States, Mexico, Australia, the United Kingdom, Germany and Poland, report that perceived Jewish power is associated with antisemitism through two distinct routes: as a threat to ingroup dominance among people who support social hierarchy, and as a threat to egalitarian ideals among people who oppose it. Harber and colleagues (2026), across three preregistered studies, report that their conspiracies-mediated model accounted for over 55% of the variance in anti-Israel attitudes, with conspiracy beliefs implicating Israel and Zionists carrying the association and a general conspiratorial mindset not explaining the result.

Where does the acronym appear in measured online data?

Zannettou and colleagues (2020) counted the word “jew” in 3.0% of 4chan /pol/ posts and 2.0% of Gab posts, according to their peer-reviewed study of over 100 million posts.

The raw counts were 1,993,432 of 67,416,903 /pol/ posts and 763,329 of 35,528,320 Gab posts, spanning July 2016 to January 2018 on /pol/ and August 2016 to January 2018 on Gab.

The same study measured the slur “kike” in 0.8% of /pol/ posts and 0.2% of Gab posts, with a 61.20-fold increase on Gab over the collection period, and found the word “jew” rising 16.44-fold between the start and end of the dataset. In random samples of 100 posts containing “jew,” 42% were judged hateful on both platforms.

McIlroy-Young and Anderson (2019) mapped Gab’s word-embedding space across posts from 73,475 active users and found the terms closest to “jew” were “kike” at 0.81, “yid” at 0.62 and “gentile” at 0.58, with antisemitic terms including “globalist” rising in frequency over the platform’s history. Bermudez-Villalva, Mehrnezhad and Toreini (2025) machine-classified about 500,000 /pol/ posts collected from May 2024 and found 11.20% classified as hateful across all categories, with antisemitism sitting inside the religious-intolerance category rather than counted separately. Bovet and Grindrod (2022), across 7 million Telegram messages from 12,564 channels between September 2015 and January 2021, recorded the UK far-right core community growing from 319 to 1,252 channels, with activity rising in mid-2019 as actors moved to Telegram after Facebook and Instagram bans.

One absence is worth recording plainly. These corpora publish shares for words and slurs, and none of them reports a measured share for the three-letter acronym itself, so no peer-reviewed prevalence figure for ZOG as a string exists in this literature.

How does a three-letter code evade moderation?

Mendelsohn, Le Bras, Choi and Sap (2023) assembled a glossary of over 300 coded terms, expressions carrying one meaning to a broad audience and a second, often hateful, meaning to an in-group, and found that harmful content containing them evaded automated toxicity detection.

Their canonical antisemitic example is “cosmopolitan,” which reads as “worldly” to most readers and as “Jewish” to a narrow in-group. The same study reports that GPT-3’s ability to decode such terms varied widely by term type and by targeted group, so a general-purpose model was not a reliable decoder.

Taylor, Peignon and Chen (2017), in a preprint, documented the mechanism directly in the 2016 “Operation Google” scheme, in which ordinary words including “skypes,” “googles” and “yahoos” were assigned an alternate hate-speech meaning, with “Skype” standing in for a Jew. Their annotators agreed more strongly about whether a word was being used for hate speech on extremist-community data, at a Cohen’s kappa of 0.871, than on keyword-collected data, at 0.676, which measures agreement about usage rather than any detection accuracy.

Arviv, Hanouna and Tsur (2021) measured the same drift in a different symbol. Searching a 10% sample of public tweets from May to June 2016 returned 803,539 tweets containing the triple-parenthesis “echo,” posted by 418,624 users, and the authors traced the symbol’s movement from marking named individuals of Jewish heritage to marking abstract targets such as “bankers” and “globalists.”

Detection work has followed the same problem. Kikkisetti and colleagues (2024), in a 659-post methodological pilot across Disqus, Telegram, Minds and GETTR seeded with the American Jewish Committee’s Translate Hate glossary and one Southern Poverty Law Center term, surfaced established coded terms including “White Genocide” and “Deep State” and flagged emerging candidates for human review; it reports no prevalence estimate. Chandra and colleagues (2021), on 3,102 labeled Twitter posts and 3,509 labeled Gab posts, reached about 91% accuracy on Gab and about 71% on Twitter for binary antisemitism detection, with combined text and image models beating text alone by about 2 to 4 percentage points on Gab.

Which older conspiracy claim does ZOG restate?

Levis Sullam, Minello, Tripodi and Warglien (2022) report that in an 1851 to 1871 sub-corpus of French periodicals and books, terms most semantically associated with “juif” skewed negative, at 399 negative against 319 positive, according to the study.

The sub-corpus comprised 26,709 periodicals and 14,465 books, and for “juifs” the researchers counted 501 negative associations against 486 positive.

The same study traces the measured economic-antisemitism stream to two named texts: Alfred Toussenel’s Les Juifs, rois de l'époque (1845), which cast the Rothschilds as the symbol of alleged Jewish financial control, and Édouard Drumont’s La France juive (1886). The narrower legend beneath that stream has a dated origin too. Georges Mathieu-Dairnvaell’s 36-page pamphlet of 1846, Histoire édifiante et curieuse de Rothschild Ier, roi des Juifs, launched the claim that Nathan Rothschild’s fortune came from advance knowledge of the 1815 Battle of Waterloo, used to manipulate the London bond market.

Niall Ferguson’s two-volume archival history The House of Rothschild (1998), written from the family’s own business archive, finds the Waterloo legend a later fabrication with no basis in the record, and establishes the family’s nineteenth-century role as large but far short of the world-controlling power alleged of it. The Institute has examined that claim and its present-day circulation separately, in a report on the Rothschild banking conspiracy.

The vocabulary persists in current corpora. Chandra and colleagues (2021) titled their multimodal detection study around the term “Jewtocracy,” a compression of the same assertion of government by a hidden group, and coded posts for category as well as presence, reaching about 67% accuracy on Gab and about 68% on Twitter on the four-class category task. Jikeli and colleagues (2022) put a share on the family of claims rather than any single term: about 62% of the antisemitic tweets mentioning “Jews” in their expert-annotated corpus fit the working definition’s paragraph on mendacious or stereotypical allegations about Jewish collective power, against 15% fitting denial of Jewish self-determination.

Is US policy toward Israel made in public?

The instruments that set US security assistance to Israel are published records. Under the Memorandum of Understanding signed at Washington on 14 September 2016, the United States pledged $38 billion over FY2019 to FY2028, per the Congressional Research Service report RL33222: $33 billion in Foreign Military Financing plus $5 billion in missile defense appropriations.

The agreement’s own text states that the funding levels “are subject to the appropriation and availability of funds,” so the document sets a level and a schedule while Congress appropriates. The appropriating decisions appear as recorded votes.

The 2021 margin sets a bounding arithmetic that no subgroup escapes: with 9 members voting no, 420 of the 431 recording a position is 97.4%, and any subgroup of 100 members voted at least 91% yes. The 140-vote difference between the 2023 and 2024 House votes on essentially the same assistance tracked the legislative packaging, since the variable between them was the IRS-rescission offset.

Advocacy spending around those votes is also filed. OpenSecrets’ tabulation of Federal Election Commission filings records $3,037,900 in direct contributions from the American Israel Public Affairs Committee’s connected PAC to federal candidates in the 2023 to 2024 cycle, 36.62% of it to Democrats and 63.02% to Republicans, spread across 347 House and 20 Senate candidates and capped by statute at $5,000 per candidate per election. The affiliated super PAC United Democracy Project reported $37,860,200 in independent expenditures in the 2024 cycle, of which $14,067,863 supported Democrats and $20,786,780 opposed Democrats. The Institute has read those filings in detail in a separate report on who sets the level of US aid to Israel.

Public opinion moved across the same period. Gallup’s February 2025 reading found 46% of US adults sympathizing more with the Israelis, the lowest in 25 years of its trend, and 33% with the Palestinians, the highest, in a sample of 1,004 adults with a margin of error of 4 points.

What does Israel’s own government report about this narrative?

Israel’s Ministry for Diaspora Affairs and Combating Antisemitism reports 124 million antisemitic posts on X during calendar 2025.

The ministry also recorded 815 antisemitic incidents, 20 Jews killed in violent incidents and 4,073 anti-Israel global protests.

These are the ministry’s own compiled counts, published by a party to the events they describe, and the ministry calls the social media figures initial findings pending a fuller report.

The incident total in that brief is media-derived, drawn from incidents documented and confirmed in mainstream and Jewish media rather than from a comprehensive register, and the social media figures come from the ministry’s own volume and sentiment analysis of one platform. Neither is an independently verified tally.

An earlier ministry publication covering 7 to 25 October 2023 reports a 500% increase in the overall volume of antisemitic events worldwide against the same period a year earlier, with a 330% increase in violent incidents, about 660% in harassments and about 128% in desecration of Jewish sites. The document itself hedges those figures with “it appears that” and “about,” publishes no baseline incident counts, names no monitoring sources and states no method, so they stand as the ministry’s own stated increases rather than measured rates.

The World Zionist Organization’s Department for Combatting Antisemitism, in its report on 2021, records an average of more than ten reported antisemitic incidents per day worldwide outside Israel, with nearly 50% occurring in Europe, compiled from its own and Jewish Agency monitoring plus local police, community organizations and research institutes. The report presents that daily average as a floor, on the stated ground that many incidents go unreported. The publication is hosted on gov.il but authored by the World Zionist Organization, an established monitoring organization with an advocacy mission rather than an Israeli ministry.

Across all three publications, one thing is absent in the same way it is absent from the peer-reviewed corpora: none of them reports a separate count for the acronym, or for the hidden-control narrative as a coded category, so a government-side figure for the term’s circulation does not exist in these records either.

Does the term spike around elections and wars?

Verma, Sear and Johnson (2024) measured antisemitic content rising 117.57% in a cross-platform hate network between 7 to 11 November 2020 and the preceding 2 to 6 November.

The network comprised about 50 million accounts, with content types classified by natural-language models that the authors report achieve 91% accuracy or higher.

The rise coincided with the US election period rather than being shown to follow from it.

The same study recorded structural change alongside the volume change: connections involving Telegram in the hate-to-hate network rose 299%, from 592 to 2,366, between 1 to 3 November and 4 to 7 November 2020, and after 6 January 2021 the network’s clustering coefficient rose 164.8% while its communities fell from 111 to 89. The authors note the new content aligned with Great Replacement narratives attributing demographic change to Jewish influence.

Sear and Johnson (2023), in a preprint mapping hate-and-extremism communities across 26 platforms, report that antisemitic hate rose sharply and almost instantaneously in the minutes following the Hamas attack of 7 October 2023, before Israel had responded, with Islamophobia rising to a much lesser extent. The same preprint describes a hate-community core of roughly 50 million individuals with direct online access to more than a billion mainstream community members, a description of network structure rather than of individual exposure.

Zannettou and colleagues (2020) recorded the same coincidence pattern in their own corpus, with antisemitic content more than doubling in some cases alongside the 2016 US presidential election and the 2017 Charlottesville rally. Against those event-linked spikes, Paudel and colleagues (2021), in a preprint tracing 189 conspiracy claims catalogued by Snopes across Reddit and Twitter, found 84.78% discussed on Reddit for longer than a year and 29.79% for more than five years, so the underlying claims persist between the spikes rather than appearing with them. The Jerusalem Post reported in 2026 that usage of ZOG surged during that year’s Iran war, a press report of a usage change rather than a measured share of a named corpus.

Does holding this belief track antisemitic attitudes?

Allington, Hirsh and Katz (2023), in a preregistered cross-sectional survey of 1,790 UK adults drawn as a YouGov quota sample in December 2021, found conspiracy suspicion correlating positively with every measured form of antisemitism, generalised, Judeophobic and anti-Zionist, with the correlation to Judeophobic antisemitism notably stronger.

Their exploratory analysis indicated the anti-Zionist link was accounted for by its overlap with the Judeophobic measure. The design is cross-sectional, so these are associations at one point in time.

Swami (2012) measured a specific conspiracy belief against ideological attitudes in two surveys of Malay adults in Kuala Lumpur, at n = 368 and n = 314, using a 12-item belief-in-Jewish-conspiracy scale. The regression model in Study 2 accounted for 39% of the variance, with attitudes toward Israel the strongest predictor at a standardized beta of 0.56, followed by modern racism toward the Chinese at 0.17, right-wing authoritarianism at 0.15 and social dominance orientation at 0.14; once ideological variables were accounted for, general conspiracist ideation was no longer a significant predictor.

The belief’s measured effects reach the people it describes. Jolley, Paterson and McNeill (2024), across three studies with Jewish participants at N = 250, n = 194 and n = 201, found that exposure to the idea that many rather than few non-Jewish people hold Jewish-targeted conspiracy beliefs increased perceived intergroup threat, which in turn raised ingroup anger and anxiety, and that higher perceived popularity made participants less likely to interact with a non-Jewish partner in a behavioural task.

Whether search-level attention to such claims relates to offline offences has been tested once at scale. Aziani, Lo Giudice and Yazdi (2025), in a preprint testing search trends for 36 racially and politically charged conspiracy theories against weekly hate-crime counts in Michigan from 2015 to 2019, found 8 of the 36 improved prediction of registered hate crimes, with the Great Replacement reducing prediction error by 3.29% at a two-week lag and 6.42% at a three-week lag, and the Rothschilds theory improving prediction in four of five model iterations. Improvements emerged two to three weeks after search fluctuations, which the authors present as a predictive relationship rather than a causal one.

The official offence count sits alongside those associations. The FBI’s Uniform Crime Reporting program recorded 1,938 anti-Jewish single-bias hate-crime incidents in 2024, the highest since collection began in 1991 and 69% of religion-based incidents, from 16,419 participating agencies covering 95.1% of the population.

Methodology and limitations

This report measures how a term and its underlying claim are defined, counted and correlated. It does not assess the assertion the acronym makes about who governs; it reports what named bodies define, what dated corpora measured, and what official and primary records document about the instruments that set US policy toward Israel.

Four source types carry the findings. Primary and official records: the 2016 Memorandum of Understanding text published by the US Department of State, the Congressional Research Service report RL33222, the recorded House and Senate votes at congress.gov, Federal Election Commission filings as tabulated by OpenSecrets, and the FBI’s 2024 hate-crime statistics. Peer-reviewed research: the corpus studies, surveys and experiments cited by author and year. Preprints, labelled at first use: Sear and Johnson (2023), Paudel and colleagues (2021), Taylor and colleagues (2017), Kikkisetti and colleagues (2024) and Aziani and colleagues (2025). Monitoring organizations with advocacy missions, whose entries are definitions rather than measurements: the Anti-Defamation League’s glossary, the American Jewish Committee’s Translate Hate entry, the World Jewish Congress conspiracy-myths catalogue and the World Zionist Organization’s 2021 report.

Government publications by a party to the events are labelled as that party’s own counts. Israel’s Ministry for Diaspora Affairs and Combating Antisemitism compiles its incident totals from media reports and its platform figures from its own volume and sentiment analysis, and calls the latter initial findings; its 2023 percentages carry no baseline counts, method or named monitoring sources. Neither set is independently verified, and neither may be summed with the FBI’s counts, which come from criminal reports under a different definition and universe.

Three limits bear on the central question. No peer-reviewed corpus study reports a measured share for the three-letter acronym itself, so no prevalence figure for the string exists in this literature; the Jerusalem Post’s 2026 report of a usage surge is a press report, not a measured share. No survey publishes a prevalence estimate for belief in this specific narrative under this specific name; the Yale Youth Poll and Hersh and Royden measure endorsement of Jewish-power statements, which is the nearest available quantity and a broader one. The FBI’s count is a known undercount: reporting is voluntary, and of the 16,419 participating agencies only 3,127, or 19.0%, reported any hate crime, while the Bureau of Justice Statistics estimated that about 42% of violent hate-crime victimizations went unreported to police between 2015 and 2019.

Machine-classified shares carry the accuracy their authors state, and detection figures for what a system can be pushed into producing are a different quantity from its default behaviour. Every association reported above is correlational, in the terms the cited authors themselves use.

Conclusion

The question begins with three letters and no footnote, and the answer to what they stand for is short: Zionist Occupation Government, defined by the Anti-Defamation League as a white supremacist acronym for the belief that the US government is controlled by Jews, dated by the American Jewish Committee to the 1970s and to movement sites including Aryan Nations and the National Alliance.

What the measurement adds is a split. The slogan sits where those definitions place it, and no corpus study reports a share for the string at all, while the claim it compresses has a broad and dated footprint: about 62% of antisemitic tweets mentioning “Jews” in Jikeli and colleagues’ annotated corpus fit the collective-power paragraph, 10% of US registered voters ages 18 to 34 in the Yale Youth Poll agreed with all three tested statements against 2% of those 65 and older, and the negative associations of “juif” outnumbered the positive ones, 399 to 319, in Levis Sullam and colleagues’ nineteenth-century French corpus. The claim also has a documented mechanism for staying in circulation: Mendelsohn and colleagues showed harmful content carrying coded terms passing automated toxicity detection, and Arviv and colleagues watched a symbol drift from named individuals to “bankers” and “globalists” across 803,539 tweets. Against the assertion of a concealed government, the instruments themselves are filed and numbered: a signed agreement setting $38 billion subject to appropriation, roll calls of 420 to 9, 226 to 196, 366 to 58 and 79 to 18, and independent expenditures itemized to the dollar.

That is the shape the data gives: a slogan of the far right, a control narrative measured across surveys and centuries, a vocabulary engineered to move quietly, and a policy record that is published line by line. The uncomfortable part is the gap between the second and the fourth. The reader who arrived at three letters is being offered an explanation of how a government works, and the sourced record shows a claim older than the term, coded to survive moderation, that surges in the minutes and days around elections and wars. Whether a compressed, unfootnoted abbreviation is one of the inputs by which that older claim reaches people who would never visit the sites where it was coined is the question the counts sharpen and do not close.

Frequently Asked Questions

Why does the acronym use “Zionist” rather than “Jewish”?

Both keyword universes have been measured, and the antisemitic share is similar in each. Jikeli and colleagues (2022) found 13.1% of tweets containing “Israel” from January to August 2020 were antisemitic under the working definition, extrapolating to about 2,392,129 tweets, against 11.2% of conversations containing “Jews” from January 2019 to August 2020. Their word-token similarity between antisemitic and non-antisemitic tweets was 0.38 for “Israel” and 0.67 for “Jews,” so the two vocabularies separate to different degrees.

How do researchers decide whether a post counts as antisemitic?

Through expert annotation against a stated definition, with the disagreement documented. Jikeli and colleagues (2022) built a gold-standard corpus of 4,016 English-language tweets annotated against the International Holocaust Remembrance Alliance definition, of which 23.1% were coded antisemitic. Their 2023 preprint extended the method to 6,941 tweets from January 2019 to December 2021 and coded 1,250, or 18%, as antisemitic.

Can automated systems detect coded antisemitism reliably?

Not uniformly, and the reported figures depend heavily on the platform and the prompt. Patel, Mehta and Blackburn (2025) tested eight open-source language models on a benchmark of 11,315 tweets, of which 1,953, or 17%, were antisemitic, and found a guided chain-of-thought prompt improved positive-class F1 by 0.03 to 0.13 over zero-shot and cut refusal rates to near 0%. Chandra and colleagues (2021) reached about 91% accuracy on Gab against about 71% on Twitter for the same binary task.

Do AI chatbots reproduce claims of hidden Jewish power?

They can be pushed into producing related content at very different rates by system. Saeed and colleagues (2024), in a preprint, measured jailbreak success in eliciting biased or stereotyped antisemitic content ranging from 0.00% for Claude 3.5 Sonnet to 76.0% for Llama 3.1-8B across six models. Dutta and colleagues (2024), in a preprint, found that of 10,484 elicited toxic expansions mentioning the Holocaust, 94.9% were assessed as misrepresenting it, including blaming Jews for it. Both measure what a system can be provoked into producing, not its default behaviour.

Does the recommendation feed carry this material to people who did not seek it?

One antisemitism-specific audit and one platform-scale experiment bear on it. The Institute for Strategic Dialogue (2025), using 10 TikTok profiles posing as 15-year-olds across over 5,500 recommended videos, reported that a profile of a boy interested in male-lifestyle-influencer content was shown antisemitic conspiracy theories within about an hour. Kalra (2025), in a preprint randomizing about 8 million users of a large short-video platform in India, found replacing engagement ranking with random content reduced toxic posts viewed by 27% while overall usage fell about 35%; that study measures anti-minority content in India, not antisemitism.

Where does the hidden-power claim appear outside far-right platforms?

In implicit vocabulary on mainstream forums. Weinberg and colleagues (2025), across over 1.26 million posts by 34,500 users in two QAnon subreddits from December 2017 to September 2018, found explicit antisemitic language in 0.66% of posts and implicit antisemitic terms in 8.6%. Of the coded sample sentences, 56% of implicit terms were tied to the hidden-power dimension of antisemitism, with the categories non-exclusive, and the authors describe their user-share figures as lower-bound proxies.

Sources

• Allington, D., Hirsh, D., and Katz, L., 2023. Correlation Between Coronavirus Conspiracism and Antisemitism: A Cross-Sectional Study in the United Kingdom. Scientific Reports. DOI 10.1038/s41598-023-41794-y. Peer-reviewed.
• American Jewish Committee. Translate Hate: Zionist Occupied Government. ajc.org. Monitoring organization (advocacy mission).
• Anti-Defamation League. Hate Symbol: ZOG. adl.org/resources/hate-symbol/zog. Monitoring organization (advocacy mission).
• Arviv, E., Hanouna, S., and Tsur, O., 2021. It’s a Thin Line Between Love and Hate: Using the Echo in Modeling Dynamics of Racist Online Communities. ICWSM 2021 (AAAI). arXiv:2012.01133. Peer-reviewed conference.
• Aziani, A., Lo Giudice, G., and Yazdi, K., 2025. Conspiracy to Commit: Information Pollution, Artificial Intelligence, and Real-World Hate Crime. arXiv:2507.07884. Preprint.
• Bermudez-Villalva, A., Mehrnezhad, M., and Toreini, E., 2025. Measuring Online Hate on 4chan using Pre-trained Deep Learning Models. IEEE Transactions on Technology and Society. DOI 10.1109/TTS.2025.3549931. Peer-reviewed.
• Bovet, A., and Grindrod, P., 2022. Organization and Evolution of the UK Far-Right Network on Telegram. Applied Network Science. DOI 10.1007/s41109-022-00513-8. Peer-reviewed.
• Bureau of Justice Statistics, 2021. Hate Crime Victimization, 2005 to 2019. National Crime Victimization Survey. bjs.ojp.gov. Official record.
• Chandra, M., Pailla, D., Bhatia, H., Sanchawala, A., Gupta, M., Shrivastava, M., and Kumaraguru, P., 2021. Subverting the Jewtocracy: Online Antisemitism Detection Using Multimodal Deep Learning. ACM Web Science Conference. arXiv:2104.05947. Peer-reviewed conference.
• Dairnvaell, G. M., 1846. Histoire édifiante et curieuse de Rothschild Ier, roi des Juifs. Paris. Primary historical artifact.
• Dutta, A., Khorramrouz, A., Dutta, S., and KhudaBukhsh, A., 2024. Down the Toxicity Rabbit Hole: A Novel Framework to Bias Audit Large Language Models. arXiv:2309.06415. Preprint.
• Federal Bureau of Investigation, 2025. Reported Crimes in the Nation, 2024 (Hate Crime Statistics). UCR Crime Data Explorer. cde.ucr.cjis.gov. Official record.
• Ferguson, N., 1998. The House of Rothschild, two volumes. Weidenfeld & Nicolson / Viking. Scholarly book (archival history).
• Gallup, 2025. Middle East sympathies trend series. news.gallup.com/poll/1639. Named survey organization.
• Hadar, B., Halevy, N., Cohen, T., Apfelbaum, E., and Chan, E., 2026. The Perils of Perceived Power: The Dual-Threat Model of Antisemitism. American Psychologist. DOI 10.1037/amp0001693. Peer-reviewed.
• Harber, K., Bulska, D., Malloy, T., and Vila, J., 2026. Antisemitism, Conspiracy Beliefs, and Anti-Israel Attitudes. American Psychologist. DOI 10.1037/amp0001635. Peer-reviewed.
• Hersh, E., and Royden, L., 2022. Antisemitic Attitudes Across the Ideological Spectrum. Political Research Quarterly. DOI 10.1177/10659129221111081. Peer-reviewed.
• Institute for Strategic Dialogue, 2025. Amplifying Antisemitism: How Recommender Algorithms Serve Harmful Content to Children. isdglobal.org. Established research organization.
• Jerusalem Post, 2026. ZOG antisemitic conspiracy theory surged during 2026 Iran war. jpost.com/diaspora/antisemitism/article-903348. Press report.
• Jikeli, G., Axelrod, D., Fischer, R., Forouzesh, E., Jeong, W., Miehling, D., and Soemer, K., 2022. Differences Between Antisemitic and Non-Antisemitic English Language Tweets. Computational and Mathematical Organization Theory. DOI 10.1007/s10588-022-09363-2. Peer-reviewed.
• Jikeli, G., Karali, S., Miehling, D., and Soemer, K., 2023. Antisemitic Messages? A Guide to High-Quality Annotation and a Labeled Dataset of Tweets. arXiv:2304.14599. Preprint.
• Jolley, D., Paterson, J., and McNeill, A., 2024. The Impact of Conspiracy Beliefs on a Targeted Group. British Journal of Psychology. DOI 10.1111/bjop.12690. Peer-reviewed.
• Kalra, A., 2025. Hate in the Time of Algorithms: Evidence on Online Behavior from a Large-Scale Experiment. arXiv:2503.06244. Preprint.
• Kikkisetti, D., Mustafa, R., Melillo, W., Corizzo, R., Boukouvalas, Z., Gill, J., and Japkowicz, N., 2024. Using LLMs to Discover Emerging Coded Antisemitic Hate-Speech in Extremist Social Media. arXiv:2401.10841. Preprint.
• Levis Sullam, S., Minello, G., Tripodi, R., and Warglien, M., 2022. Representation of Jews and Anti-Jewish Bias in 19th Century French Public Discourse. Frontiers in Big Data. DOI 10.3389/fdata.2021.723043. Peer-reviewed.
• McIlroy-Young, R., and Anderson, A., 2019. From “Welcome New Gabbers” to the Pittsburgh Synagogue Shooting: The Evolution of Gab. arXiv:1912.11278 (AAAI). Peer-reviewed conference.
• Mendelsohn, J., Le Bras, R., Choi, Y., and Sap, M., 2023. From Dogwhistles to Bullhorns: Unveiling Coded Rhetoric with Language Models. ACL 2023. arXiv:2305.17174. Peer-reviewed conference.
• Ministry for Diaspora Affairs and Combating Antisemitism, 2023. The State of Antisemitism in the World: A Summary of the First Three Weeks of the “Swords of Iron” War. gov.il. Official record (Israeli government body, self-reported).
• Ministry for Diaspora Affairs and Combating Antisemitism, 2026. Annual Report Executive Brief: Global Overview of Antisemitism in 2025. gov.il. Official record (Israeli government body, self-reported).
• OpenSecrets, 2025. AIPAC PAC contributions to federal candidates, 2023 to 2024 cycle (FEC committee C00797670). opensecrets.org. Campaign-finance aggregator of FEC filings.
• OpenSecrets, 2025. United Democracy Project independent expenditures, 2024 cycle (FEC committee C00799031). opensecrets.org. Campaign-finance aggregator of FEC filings.
• Patel, J., Mehta, S., and Blackburn, J., 2025. Evaluating Large Language Models for Detecting Antisemitism. EMNLP 2025. arXiv:2509.18293. Peer-reviewed conference.
• Paudel, P., Blackburn, J., De Cristofaro, E., Zannettou, S., and Stringhini, G., 2021. Soros, Child Sacrifices, and 5G: Understanding the Spread of Conspiracy Theories on Web Communities. arXiv:2111.02187. Preprint.
• Sear, R., and Johnson, N., 2023. Unprecedented Reach and Rich Online Journeys Drive Hate and Extremism Globally. arXiv:2311.08258. Preprint.
• Swami, V., 2012. Social Psychological Origins of Conspiracy Theories: The Case of the Jewish Conspiracy Theory in Malaysia. Frontiers in Psychology. DOI 10.3389/fpsyg.2012.00280. Peer-reviewed.
• Taylor, J., Peignon, M., and Chen, Y.-S., 2017. Surfacing Contextual Hate Speech Words within Social Media. arXiv:1711.10093. Preprint.
• US Congress, 2021 to 2024. Recorded votes on Israel security assistance: House Roll Calls 275, 577 and 152; Senate Record Vote 154. congress.gov. Official record (recorded votes).
• US Congressional Research Service. U.S. Foreign Aid to Israel: Overview and Developments (RL33222). congress.gov. Official record.
• US Department of State, 2016. Memorandum of Understanding Between the United States and Israel on Security Assistance. 2009-2017.state.gov. Official record (bilateral executive agreement).
• Verma, A., Sear, R., and Johnson, N., 2024. How U.S. Presidential Elections Strengthen Global Hate Networks. npj Complexity. DOI 10.1038/s44260-024-00018-8. Peer-reviewed.
• Weinberg, D., Levy, A., Edwards, T., Kopstein, J., Frey, W., and colleagues, 2025. Hidden in Plain Sight: Antisemitic Content in QAnon Subreddits. PLoS ONE. DOI 10.1371/journal.pone.0318988. Peer-reviewed.
• World Jewish Congress. Conspiracy Myths: Zionist Occupied Government (ZOG). worldjewishcongress.org. Monitoring organization (advocacy mission).
• World Zionist Organization, Department for Combatting Antisemitism and Enhancing Resilience, 2022. The State of Antisemitism in 2021. gov.il. Monitoring organization (advocacy mission).
• Yale Youth Poll, 2025. Fall 2025 Poll on Antisemitic Attitudes and Israel-Palestine. youthpoll.yale.edu. Academic poll.
• Zannettou, S., Finkelstein, J., Bradlyn, B., and Blackburn, J., 2020. A Quantitative Approach to Understanding Online Antisemitism. ICWSM 2020 (AAAI). DOI 10.1609/icwsm.v14i1.7343. Peer-reviewed conference.