The Surge of Anti-Semitism in German Social Media following the October 7 Attacks

Gregor Wiedemann; Daniel Wehrend

cs.SISep 28, 2026

The Surge of Anti-Semitism in German Social Media following the October 7 Attacks

Authors: Gregor Wiedemann, Daniel Wehrend

Organizations: Leibniz-Institute for Media Research | Hans-Bredow-Institut Hamburg, Germany

Abstract

We investigate the extent to which the Hamas attacks on Israel of October 7, 2023, have affected German social media debates about Judaism and Israel. For this, we develop an approach to detect 26 anti-Semitic categories in user postings via large language models (LLMs). The approach is applied to Facebook and Telegram posts (N=125,718) from three months before and after the event. Methodically, we test different open-weight models in two setups---with and without user information as additional context to the post text. The best setup achieves up to 83 % F1-score for binary anti-Semitism detection on our manually coded validation set. User context provides valuable information for most LLMs and drastically reduces false positives, for example, when (critically) reporting on anti-Semitic incidents. Concerning our topic, we find that anti-Semitism is surging significantly on both platforms, while being about ten times more prevalent on Telegram compared to Facebook. Facebook users express anti-Semitic views most likely in posts about an alleged genocide in Gaza carried out by the Israeli army, whereas classic anti-Semitic stereotypes related to power and conspiracy theories are dominant on Telegram. After the attack, the discourse patterns on both platforms show signs of convergence, as classic anti-Semitism increases on Facebook, whereas Israel-related categories surge on Telegram.

Figures & tables

Platform	Facebook	Telegram
N posts	58,411	67,307
Share pre Oct. 7	21.2 %	9.6 %
Share post Oct. 7	78.8 %	90.4 %
Avg. text length (chars)	622	649
N users/channels	12,577	1,000
Avg. posts per user/channel	4.64	67.31

Table 1: Dataset statistics by platform

Figure 1: Post frequency per platform over time. The vertical line marks October 7, 2023.

Model	No user context				With user context
	P	R	F	A	P	R	F	A
DeepSeek-r1 14b DeepSeek-AI et al. (2025)	0.68	0.64	0.66	0.76	0.71	0.67	0.69	0.78
DeepSeek-r1 32b DeepSeek-AI et al. (2025)	0.74	0.54	0.62	0.76	0.77	0.59	0.67	0.79
DeepSeek-r1 70b DeepSeek-AI et al. (2025)	0.76	0.72	0.74	0.82	0.78	0.77	0.78	0.84
Gemma 3 27b Gemma et al. (2025)	0.64	0.92	0.76	0.78	0.62	0.94	0.75	0.77
Gemma 4 26b Gemma 4 Team (2026)	0.92	0.48	0.63	0.80	0.90	0.44	0.59	0.78

Table 2: Text classification performance for binary antisemitism detection in two experiment setups—with and without user information as additional context (precision, recall, F1-score, and accuracy for the is_antisemitic=True minority class).

Figure 2: Frequency of anti-Semitic categories per platform

Figure 3: Share of anti-Semitic categories per platform

Class	P	R	F1	N	Acc.
K20	0.87	0.76	0.81	17	0.993
K15	0.79	0.79	0.79	14	0.993
K13	0.88	0.65	0.75	23	0.988
K07	0.77	0.72	0.74	121	0.925
K17	0.69	0.75	0.72	61	0.955
K11	0.67	0.76	0.71	21	0.984

Table 3: Text classification performance of gpt-oss 120b for individual categories (precision, recall, F1-score, and accuracy; N: number of ‘positive’ test instances)

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Code	Name (translation)	Short definition (translated from German)	N	$\mathbf{\kappa}$
K01	The Other / Foreigner	Jews are constructed as fundamentally different or not belonging, a foreign out-group that does not really belong to the national or social community.	6	0.26
K02	Evil	Attribution of an ontological, global evil nature to Jews, Israelis, Israel or Zionism, as if they had an inherently evil core or destructive essence.	31	0.38
K03	Blood libel / Child murder	Claims that Jews or Israelis deliberately kill non-Jewish children or use their blood, often framed as ritual or systematic killing without evidence.	10	0.57
K04	Immorality	Jews or Israelis are portrayed as fundamentally without conscience or empathy, often combined with accusations of hypocrisy or instrumentalization of the Holocaust.	21	0.32
K05	Lies and deceit	Jews or Israelis are depicted as inherently dishonest and manipulative, with generalized accusations of systematic lying, distortion and deception.	14	0.30
K06	Blaming Jews for antisemitism	Responsibility for antisemitic incidents or hostility is shifted onto Jews themselves, reversing victim–perpetrator roles by claiming that Jewish behaviour causes antisemitism.	1	0.00

Appendix

Table 4: Categories of antisemitism, short definitions, number of examples, and Cohen’s Kappa inter-coder agreement per category in the expert validation set. Names and descriptions are translated from the German original data. For the full, original codebook and corresponding LLM prompts, see the online resources (Appendix B ).

Figure 4: Label co-occurrence heatmap of true (rows) and predicted (columns) labels (co-occurrence counts show which true labels tend to get predicted together).

Figure 5: Difference of category shares pre/post the October 7 attacks with regard to the overall number of posts on the platform.

Social Media Analysis

Social media analysis studies online posts, media, users, and interactions to understand communication, communities, and public behavior.

See all 221 papers

Evaluating Large Language Models for Antisemitic Incident Classification
Jul 6, 2026Karina Halevy, Julia Mendelsohn, Chan Young Park +2LLM Evaluation Hate Speech Detection
Carnegie Mellon University · University of Maryland · Microsoft Research +2
You Frame It: How Conceptual Representations Shape LLM Detection and Reasoning about Antisemitism
Jul 6, 2026Katharina Soemer, Helena MihaljevićLLM Grounding Hate Speech Detection
Goethe University, Frankfurt, Germany · HTW Berlin, Germany
Leveraging LLMs for Context-Aware Implicit Textual and Multimodal Hate Speech Detection
Oct 17, 2025Joshua Wolfe Brook, Ilia MarkovHate Speech Detection
Computational Linguistics & Text Mining Lab, Vrije Universiteit Amsterdam

Jul 6, 2026J/K move · Enter open · S save

Evaluating Large Language Models for Antisemitic Incident Classification

Karina Halevy, Julia Mendelsohn, Chan Young Park, Yulia Tsvetkov+1

Carnegie Mellon University · University of Maryland · Microsoft Research+2

LLM Evaluation Hate Speech Detection

Open paper AskPDF arXiv

Addressing hate and violence in society requires timely detection of hateful events from public reporting, but automated identification of hateful events remains underexplored. We introduce the task of hateful event detection and investigate the ability of AI systems, specifically large language models (LLMs), to discover and classify reports of antisemitic events with fine-grained labels. We evaluate OpenAI's GPT-4o and Meta's Llama-3.2-3B-Instruct on multiple expert-annotated datasets containing antisemitic event descriptions from news articles, civil society reports, and official records. We show that LLMs, particularly GPT-4o, have potential for this task, but substantial improvement is needed. Providing clear term definitions and in-context examples in prompts can improve performance: definitions are most helpful for rhetoric-oriented events (e.g. classical antisemitic tropes), while examples help label action-oriented events (e.g. physical assault). A case study of college newspapers demonstrates that LLMs can help surface relevant real-world events, supporting early monitoring and intervention. Overall, our findings highlight both opportunities and critical gaps in AI's ability to recognize complex harms and underscore the need for collaborative efforts among AI developers, policymakers, and civil society to design models, implement robust evaluation, and develop policy frameworks for defining and combating hate efficiently and effectively.

Abstract

Social Media Analysis

Explore similar work

Evaluating Large Language Models for Antisemitic Incident Classification

You Frame It: How Conceptual Representations Shape LLM Detection and Reasoning about Antisemitism

Leveraging LLMs for Context-Aware Implicit Textual and Multimodal Hate Speech Detection