Monthly Report August 2025

Data for August 2025 was based on 10.2M messages in 25 languages across 12 social media platforms such as X, YouTube, TikTok, Facebook, and Threads.

Content warning: Presented data may contain disturbing language related to online hate speech.

Average toxicity

In August, the overall average toxicity level was 0.19. The month began with lower values of 0.17, but toxicity gradually increased to 0.18-0.19 by mid-August. Brief peaks of 0.20 appeared on the 17th, 21st-23rd, and 27th, although these were short-lived. Overall, the trend shows a steady rise from the start of the month, levelling out at around 0.19 in the final week.

Baseline channel analysis

Throughout August, toxicity levels across the monitored baselines remained largely consistent, with some notable fluctuations. The antisemitism baseline stood out as the most persistently toxic, ranging between 0.30 and 0.32 throughout the month, maintaining its position as the highest among all narratives. Anti-LGBTQ+ discourse followed closely, with levels hovering between 0.26 and 0.28, showing steady but elevated toxicity. Anti-Muslim toxicity was slightly lower, fluctuating between 0.23 and 0.26, with small peaks in mid- and late August. In contrast, sexism remained the least toxic category, averaging between 0.14 and 0.17, although a mild upward trend was visible in the second half of the month. Anti-refugee narratives showed little variation, with a stable average toxicity of 0.22, peaking at 0.25 in the second half of the month. At the same time, the anti-Roma baseline displayed the highest volatility: values fluctuated sharply from 0.16 to 0.23, reaching the monthly average on 30 August. It is important to note that the Roma baseline has the smallest sample size (180.6K), which makes the average toxicity more susceptible to fluctuations.

Baseline Sexism

Volume: 577.4k of 6.4M posts (9.0%) in this channel were identified as toxic content.

Platforms: Most data was collected from YouTube (49%) and X (42%).

Languages: Posts were mostly written in English (60%).

Toxicity: The average toxicity score was 0.16. Posts with very high toxicity (≥0.8) totaled 29k, and were found predominantly on YouTube (49%).

Themes: 

• 53% of toxic posts were categorized as Sexism.

• 36% of those toxic posts contained violent rhetoric.

Common toxic words included: women, quran, woman, fat, الله (Allāh), ser, اليهود (jew)  slayer, femme

Key Insights: A recurring theme is the use of derogatory and vulgar language directed at women, frequently employing sexual insults and dehumanising terms. Many posts express intense anger and contempt towards men, accusing them of being unintelligent, incapable of listening, and generally useless. There's also a significant amount of xenophobic and racist sentiment, particularly targeting specific ethnic or national groups with slurs and accusations of criminality. 

Selected news items:

• Eurodeputata e consigliera vittime del forum sessista - Notizie - Ansa.it

• Il Pd: 'Azione bipartisan anti-sessismo'. Si muove il governo - Notizie - Ansa.it

• Horváth László drogügyi kormánybiztos megvédte a drogozásról, kiskorúak megerőszakolásról és nők veréséről rappelő Dopemant

• Lilly Téllez: criatura de madera   | Opinión

• Donald Trumps Washington: Ein Sumpf, den niemand mehr trockenlegt

Baseline Anti-LGBTQ+

Volume: 109.2k of 761.4k posts (14.3%) in this channel were identified as toxic content.

Platforms: Most data was collected from X (69%) and YouTube (24%).

Languages: Posts were mostly written in English (61%).

Toxicity: The average toxicity score was 0.27. Posts with very high toxicity (≥0.8) totaled 5k, and were found predominantly on X (69%).

Themes: 

• 73% of toxic posts were categorized as Sexism.

• 20% of those toxic posts contained violent rhetoric.

Common toxic posts included: gay, شاذ (shādh), lesbian, homo, ser, sexy, sex, women, porn, schwule

Key Insights: A recurring theme is the association of LGBTQ+ people with pedophilia and degeneracy, a harmful stereotype used to dehumanise and condemn them. Many posts employ derogatory and vulgar language, employing slurs and insults to express extreme disapproval and hatred. Additionally, the posts frequently conflate LGBTQ+ identities with other marginalised groups, such as racial or ethnic minorities, and often incorporate conspiracy theories or hateful ideologies. The language used is consistently aggressive, dehumanising, and it aims to incite disgust and revulsion towards the targeted communities.

Selected news items:

• Fall Marla Svenja Liebich: Queerbeauftragte warnt vor rechter Stimmungsmache gegen trans Personen

• Budapester Bürgermeister wegen Pride-Parade von der Polizei vernommen

• Sachsen: CSD in Bautzen verläuft ruhig - 3.000 Menschen demonstrieren

• Saksassa kuohuu: Viharikoksista tuomittu uusnatsi muutti sukupuolensa - passitettiin naisvankilaan

• Detaily masakru v katolické škole: Vraždil bývalý transgender žák. Znepokojivá videa na YouTube!

Baseline Anti-Muslim

Volume: 231.4k of 1.3M posts (18.0%) in this channel were identified as toxic content.

Platforms: Most data was collected from X (65%) and YouTube (29%).

Languages: Posts were mostly written in English (51%) and Arabic (16%).

Toxicity: The average toxicity score was 0.25. Posts with very high toxicity (≥0.8) totaled 23k, and were found predominantly on X (65%).

Themes: 

• 83% of toxic posts were categorized as Religion.

• 39% of those toxic posts contained violent rhetoric.

Common toxic words included: quran, muslim, islamic, muslims, الله (Allāh), moros, islam, moro, الإسلام (al-islām), israel

Key Insights: A recurring theme is the portrayal of Muslims as inherently violent, backward, and a threat to Western societies. Some posts connect anti-Muslim sentiment with broader political ideologies, such as nationalism and anti-immigration stances. The content frequently resorts to conspiracy theories and misinformation to fuel the negative narrative. Overall, the posts reveal a pattern of demonisation, prejudice, and calls for hostility towards Muslims.

Selected news items:

• PP y Vox prohíben en Jumilla la celebración de dos grandes festividades musulmanas en espacios públicos

• El presidente de la Federación de Entidades Islámicas, sobre Jumilla: "Es todo por un puñado de votos y a costa del miedo de los ciudadanos"

• Vox convierte la secularización en herramienta política contra el islam: "Esto afecta a todas las religiones"

• Eva Gregersen: Samira Nawas islamvenlige politik er undergravende for det liberale demokrati | Berlingske

• Miras reforzará con 700 agentes las policías locales de su comunidad, tras las cacerías racistas de Torre Pacheco

Baseline Anti-Refugee/Migrants

Volume: 51.7k of 416.9k posts (12.4%) in this channel were identified as toxic content.

Platforms: Most data was collected from X (70%) and YouTube (21%).

Languages: Posts were mostly written in English (42%), Spanish (14%), and German (11%).

Toxicity: The average toxicity score was 0.23. Posts with very high toxicity (≥0.8) totaled 2k, and were found predominantly on X (70%).

Themes: 

• 70% of toxic posts were categorized as Politics.

• 40% of those toxic posts contained violent rhetoric.

Common toxic keywords included: israel, migrant, ser, refugees, israël, deutschland, asielzoeker, لاجئ (lājiʾ), immigrati, asylanten

Key Insights: A recurring theme is the perception that refugees and migrants are not genuine asylum seekers but rather "invaders" or "scum" who pose a threat to the host country's culture, safety, and economy. There's a notable distrust of authorities and media, accused of being biased, corrupt, or deliberately downplaying the negative impacts of immigration. Some posts advocate for harsh measures, including deportation and even violence, against migrants. Conversely, a few posts defend migrants, highlighting their contributions or criticizing the xenophobic rhetoric. The discourse is highly polarised, with strong emotions and accusations of racism and fascism being exchanged.

Selected news items: 

• El portavoz de Vox en las Corts Valencianes compara la "inseguridad" por la inmigración con los atentados de ETA

• Cientos de inmigrantes rodean una comisaría en Montornés del Vallés al grito de "¡Policía asesina!"

• " Soros et sa bande de psychopathes ont causé de graves dommages " : Trump menace d'attaquer en justice le milliardaire philanthrope

• Donald Trump menace de poursuivre en justice le milliardaire George Soros

• Nemecko predĺži hraničné kontroly, chce aj deportovať viac ľudí

Baseline Antisemitism

Volume: 241.7k of 1.1M posts (21.3%) in this channel were identified as toxic content.

Platforms: Most data was collected from X (76%) and YouTube (17%).

Languages: Posts were mostly written in English (58%).

Toxicity: The average toxicity score was 0.31. Posts with very high toxicity (≥0.8) totaled 19k, and were found predominantly on X (76%).

Themes: 

• 85% of toxic posts were categorized as Racism.

• 40% of those toxic posts contained violent rhetoric.

Common toxic keywords included: jews, israel, jewish, اليهود (al-yahūd), israël, judíos, juifs, jew, żydów, يهودي (yahūdī)

Key Insights: The posts reveal a strong undercurrent of antisemitism, dominated by conspiracy theories portraying Jews as controlling global systems. Another significant theme is the denial or distortion of the Holocaust, with some posts questioning its legitimacy or minimising its scale. Posts also feature dehumanising language, violent rhetoric, and links to other forms of hate, such as Islamophobia and racism, showing how antisemitism intersects with broader extremist narratives.

Selected news items:

• Rhône : la plaque en hommage aux Justes de Villeurbanne arrachée, les élus dénoncent un " un geste lâche

• Pour la sœur d'Ilan Halimi, Emmanuel Macron "ne fait rien" pour lutter contre l'antisémitisme

• Herdenkingsboom voor doodgemartelde jood geveld nabij Parijs

• Los terroristas de Hamás obligan a cavar su propia tumba a un rehén judío consumido por el hambre

• Un olivier rendant hommage à Ilan Halimi, séquestré et battu à mort en 2006, a été abattu à Épinay-sur-Seine

Baseline Roma

Volume: 28.7k of 180.6k posts (15.9%) in this channel were identified as toxic content.

Platforms: Most data was collected from X (49%) and YouTube (43%).

Languages: Posts were mostly written in Russian (25%), English (20%), and Portuguese (16%).

Toxicity: The average toxicity score was 0.21. Posts with very high toxicity (≥0.8) totaled 2k, and were found predominantly on X (49%).

Themes: 

• 87% of toxic posts were categorized as Racism.

• 15% of those toxic posts contained violent rhetoric.

Common toxic keywords included: gitanos, gitano, gypsies, gypsy, zigeuner, roms, ser, moros, cygan, jews

Key Insights: The posts show widespread anti-Roma sentiment, dominated by stereotypes linking Roma people to crime, deceit, and societal burdens. Language is often dehumanising and exclusionary, with calls for expulsion, frequent use of slurs, and generalisations about the entire group. Some posts touch upon historical persecution, referencing the Holocaust, while others use derogatory terms and slurs.

Selected news items:

• Navracsics Tibor visszautasítja, hogy romaellenes lenne az önazonossági törvény

• Orbán Viktor egy DESH-nótával üzent a cigányoknak

• " Ils viennent toujours dans les mêmes communes ! " : le bourgmestre d'Enghien réagit après le forcing des gens du voyage, ce week-end

Sexism:

• Highest share of neutral messages (61.4%).

Anti-Roma:

• Neutral messages make up 54.3%.

• Medium-toxicity messages account for 24.5%.

Anti-Muslim:

• Neutral share at 44.57%.

• Medium-toxicity content at 25.17%.

• Highest proportion of high-toxicity content (1.78%).

Anti-LGBTQ+:

• Lowest neutral share (29.7%), indicating highly polarised discussions.

• Highest share of low-toxicity content (45.47%).

Antisemitism:

• Neutral share at 30.81%, also indicating polarised discourse.

• Medium-toxicity content at 33.64%.

• High-toxicity content at 1.66% (second highest after anti-Muslim).

Hate speech by category

The hate speech categorisation data reveals clear patterns across the six monitored baselines: antisemitic content is overwhelmingly framed through racism (81.41%) and religion (47.10%), with notable levels of threat (20.67%). Anti-Muslim discourse is dominated by religious framing (76.94%) and racism (48.49%), marking it as one of the most intersectional baselines. Anti-LGBTQ+ content strongly overlaps with sexism (60.91%) and obscenity (8.93%), reflecting its focus on gender and identity-based abuse. Anti-refugee narratives are primarily political (49.02%) and racist (41.67%), often portraying refugees as a societal threat. Anti-Roma content is also heavily racialised (39.82%), though less dominated by politics or religion than other baselines. Sexism-related discourse intersects with sexism itself (38.48%) and ridicule (10.56%). Across all baselines, racism, politics, religion, and sexism emerge as the key vehicles of hate, with antisemitic and anti-Muslim content facing the highest levels of violent threats.