Monthly Report July 2025
Data for July 2025 was based on 8.4M messages in 26 languages across 12 social media platforms such as YouTube, X, TikTok, Threads, and Instagram.
Content warning: Presented data may contain disturbing language related to online hate speech.
Average toxicity
In July, the overall average toxicity level was 0.18. The month started slightly higher, with values of 0.19-0.20 during the first week, before gradually declining. By mid-July, toxicity dropped to 0.17 and remained consistently low for the majority of the month. The lowest point was recorded on 29 July at 0.15, after which levels stabilised around 0.16-0.17. Overall, the trend reflects a steady decrease from the beginning of the month, with a period of prolonged stability at lower levels.
Baseline channel analysis
In July, baseline toxicity levels displayed both stability and moments of fluctuation across narratives. Antisemitism remained the most persistently toxic, holding steady at around 0.29-0.30 for nearly the entire month, only dipping slightly to 0.28 on the final day. Anti-LGBTQ+ discourse followed a similar pattern of consistency, with values hovering between 0.26 and 0.27 throughout.
Anti-Muslim content showed slightly more variation, ranging from 0.21 to 0.27, with an uptick between 12th and 13th of July before returning to lower averages. Sexism consistently recorded the lowest toxicity, maintaining a trend between 0.14 and 0.16. Anti-refugee narratives were also relatively stable, averaging between 0.20 and 0.22, with minor increases mid-month.
By contrast, the anti-Roma baseline displayed the highest volatility. While values often aligned with other narratives at around 0.25-0.27, they spiked sharply to 0.34 on July 12th before declining to 0.19 on July 16th and stabilising around 0.24-0.26 toward the end of the month.
Baseline Sexism
Volume: 2.3M of 25.7M posts (9.0%) in this channel were identified as toxic.
Platforms: Most data was observed on X (51%) and YouTube (31%).
Languages: Posts are mostly written in English (53%).
Toxicity Analysis: The average toxicity score was 0.16. Posts with very high toxicity (≥0.8) totalled 98k, and are most prominent on Gab (2%).
Toxicity Categories:
52% of toxic posts were categorised as Sexism.
32% of those toxic posts contained violent rhetoric.
Commonly used words included: women, woman, quran, الله, girl, femmes, stupid, femme, homme, vrouwen
Key Findings: Several posts contain explicit misogynistic slurs and threats, with women being referred to with terms like "pig" "sow" "witch" and "whore". There are also instances of racial slurs and dehumanising language used against various ethnic groups. Political figures and groups are frequently targeted with insults, with terms like "terrorist state Israel," "Zionists," "fascists," and "communists" appearing. The content also includes accusations of criminal activity, violence, and sexual assault. Some posts express extreme anger and calls for violence, while others engage in political commentary and criticism.
Selected News items
• A poklot járta meg a magyar származású lány: így zaklatták a Hamasz túszejtői
• Santurtzi se planta ante las agresiones machistas | El Correo
• Missbrauch im Schwimmbad - Opfer (16) bricht Schweigen | Heute.at
• La Policía Local de Toledo se suma al sistema estatal de seguimiento de casos de violencia de género
Baseline Anti-LGBTQ+
Volume: 548k of 3.6M posts (15.1%) in this channel were identified as toxic.
Platforms: Most data was observed on X (73%) and YouTube (13%).
Languages: Posts are mostly written in English (48%) and Spanish (12%).
Toxicity: The average toxicity score was 0.27. Posts with very high toxicity (≥0.8) totalled 22k, and are most prominent on Gab (4%).
Toxicity:
74% of toxic posts were categorised as Sexism.
14% of those toxic posts contained violent rhetoric.
Commonly used words included: homo, women, schwule, lésbica, lesbian, lesbica, lesbiana, lesben, queer, travesti
Key Findings: The provided posts contain a significant amount of hateful and derogatory language directed towards LGBTQ+ individuals. Several posts from different European countries, including France, Germany, the Netherlands, Poland, and Spain, use slurs and express extreme animosity. Some posts specifically target gay men, lesbians, and transgender individuals, often associating them with negative stereotypes or criminal behaviour. For instance, one post from the Netherlands refers to individuals as "faggots" and "imbeciles" while another from Germany uses terms like "queer pissers" and "queer scum". A post in Spanish from Spain uses highly offensive language towards a transgender politician. There are also instances of religious condemnation, with one post in Arabic referencing "infidels" and "heretics."
Selected News items
• U baptističkoj crkvi u Americi molili za pogubljenja LGBT osoba
• Reform UK run council removes all trans books from children's library section
• Bundestag lässt Regenbogenflaggen aus Abgeordnetenbüros entfernen
• Nach Pride-Parade: Budapester Polizei lädt Bürgermeister als Verdächtigen zu Befragung vor
Baseline Anti-Muslim
Volume: 1.1M of 6.1M posts (17.7%) in this channel were identified as toxic.
Platforms: Most data was observed on X (74%), and YouTube (17%).
Languages: Posts are mostly written in English (43%), Arabic (17%), and French (12%).
Toxicity: The average toxicity score was 0.25. Posts with very high toxicity (≥0.8) totalled 90k, and are most prominent on Gab (3%).
Toxicity Categories:
79% of toxic posts were categorised as Religion.
33% of those toxic posts contained violent rhetoric.
Commonly used words included: moros, muslim, muslims, moro, الله, islamic, quran, islam, الإسلام (Islam), المسلمين (Muslim)
Key Findings: The social media posts reveal a strong undercurrent of anti-Muslim sentiment expressed through derogatory language, accusations of violence and cultural imposition, and calls for exclusion. Several posts specifically mention "moros" (Moors) in Spain, linking them to crime, violence, and the perceived "selling out" of the country by political parties. In Germany, similar sentiments are voiced, with one post referencing "Ziegenficker" (goat fucker) and accusing Muslims of lying and deceit, referencing a "Surre" (likely a misspelling of Surah). French posts also contain harsh rhetoric, with terms like "crevures islamiques" (Islamic vermin) and calls for their eradication. Arabic posts echo these sentiments, criticizing "Al-Majd" (glory) and associating it with "Al-Khawarij" (a historical dissident group) and "Al-Ansar" (supporters of the Prophet Muhammad). There are also mentions of specific events or figures, such as "Hamas-Nazis" and comparisons to Hitler, indicating a conflation of Islam with extremist ideologies. The language used is consistently aggressive and dehumanizing, employing slurs and violent imagery to express animosity towards Muslims.
Selected News items
• Las federaciones musulmanas de Cataluña condenan los "ataques cobardes" a la mezquita de Piera
• Manifestación unitaria contra el odio y el racismo en Piera
Baseline Anti-Refugee/Migrants
Volume: 246.5k of 2.2M posts (11.5%) in this channel were identified as toxic.
Platforms: Most data was observed on X (76%) and YouTube (12%).
Languages: Posts are mostly written in English (34%), Spanish (16%), and German (13%).
Toxicity: The average toxicity score was 0.22. Posts with very high toxicity (≥0.8) totalled 10k, and are most prominent on Gab (1%).
Themes:
67% of toxic posts were categorised as Politics.
33% of those toxic posts contained violent rhetoric.
Commonly used words included: refugees, migrant, لاجئ (refugee), migrante, asylanten, israel, immigrati, deutschland, vluchtelingen, israël
Key Findings: The provided posts reveal a strong undercurrent of anti-immigrant and anti-refugee sentiment across various European countries. Several posts express anger and frustration towards immigrants, often associating them with crime, such as "Messerstecher und Vergewaltiger" (knife-wielding attackers and rapists) in Germany. The term "Sozialschmarotzer" (social parasites) is frequently used to describe immigrants who are perceived as exploiting welfare systems. There's a recurring theme of blaming immigrants for societal problems and a perceived decline in national identity and values. Some posts specifically mention "Syrians" and "Afghans" as unwelcome groups. In Spain, there are mentions of "inmigrantes clandestinos" (illegal immigrants) and a desire for their expulsion. The sentiment extends to criticism of political leaders and media outlets, accused of being too lenient or biased towards immigrants. For instance, one post criticises "unsere verblödete Regierung und der linksgrüne Misthaufen" (our brain-dead government and the left-green dung heap) in Germany for allegedly favoring "Invasoren" (invaders). The posts also highlight a sense of betrayal and a feeling that native populations are being disregarded or even harmed by current policies. The language used is often aggressive and dehumanising, employing terms like "Gesindel" (scum), "Pack" (rabble), and "Abschaum" (scum/dregs).
Selected News items
• Migránsokról, agyhalottakról
• El Gobierno, los populares y partidos de izquierda rechazan la deportación masiva que propone Vox
• Detectan en Vitoria a un primer grupo de una docena de refugiados de Malí pernoctando en la calle
Baseline Antisemitism
Volume: 1.1M of 5.6M posts (20.4%) in this channel were identified as toxic.
Platforms: Most data was observed on X (77%), and YouTube (12%).
Languages: Posts are mostly written in English (54%).
Toxicity Analysis: The average toxicity score was 0.30. Posts with very high toxicity (≥0.8) totalled 77k, and are most prominent on Gab (5%).
Themes:
82% of toxic posts were categorised as Racism.
36% of those toxic posts contained violent rhetoric.
Commonly used words included: jews, اليهود, jewish, israel, judíos, juifs, israël, يهودي, żydów, juif
Key Findings: Recent social media posts reveal a concerning trend of hateful rhetoric and conspiracy theories targeting Jewish people. Several posts, originating from various European countries, employ derogatory terms and perpetuate harmful stereotypes. For instance, one post from Germany, referencing a demonstration, quotes the chant "Kill Jews." Another post from an unspecified European location directly addresses "Zionists," labeling them as "nothing more than Satan worshippers." There are also instances of historical revisionism, with one post claiming that "Jews are the only people in the world who are 'expected' to feed and water those who are trying to murder Jews." The posts also frequently link Jewish people to negative global events and control, such as one claiming "The Jew is rattling the Goyim's chain" and another stating "American government is fully infiltrated and controlled by the parasitic jews." The language used is often aggressive and dehumanizing, with some posts explicitly calling for violence. For example, one post states, "Jews are the only people in the world who are 'expected' to feed and water those who are trying to murder Jews." Another post from Poland directly states, "JEBAĆ ŻYDÓW" (Fuck Jews). These sentiments are not isolated, as similar hateful messages appear across multiple posts, indicating a widespread issue of online hate speech directed at Jewish individuals and communities.
Selected News items
• Elon Musk chatbotja MechaHitler névre keresztelte magát, és ismét zsidózásba kezdett
• Elmo gehackt: Antisemitische Posts auf Trumps X-Account
• Macron announces Dreyfus memorial day amid rising anti-Semitism
Baseline Roma
Volume: 154.5k of 731.2k posts (21.1%) in this channel were identified as toxic.
Platforms: Most data was observed on X (64%) and YouTube (24%).
Languages: Posts are mostly written in English (22%), Spanish (15%), Russian (15%), and Portuguese (14%).
Toxicity Analysis: The average toxicity score was 0.24. Posts with very high toxicity (≥0.8) totalled 10k, and are most prominent on Gab (6%).
Themes:
88% of toxic posts were categorised as Racism.
12% of those toxic posts contained violent rhetoric.
Commonly used words included: gitanos, gitano, gypsies, zigeuner, roms, gypsy, moros, cygan, jews, roma
Key Findings: The provided social media posts reveal a complex and often contentious discourse surrounding various ethnic and social groups. Several posts directly use derogatory terms such as "Zigeuner" (gypsy), "Neger" (negro), and "Kanaken" (a derogatory term for foreigners, particularly those of Turkish or Middle Eastern descent). These terms are frequently employed in contexts of insult, complaint, or to express negative stereotypes. In Spain, a post questions whether one can call a "gitano" a "gitano" and a "negro" a "negro," but not in a pejorative way, linking it to respect and the potential for offense. Conversely, other posts highlight the perceived negative behaviors associated with certain groups, such as "Taschendiebstahl- Zigeuner" (pickpocketing gypsies) or "Syrischen u Arabischstämmigen Vergewaltiger und Messerstecher" (Syrian and Arab-origin rapists and stabbers). There are also instances where these terms are used in a seemingly neutral or even self-identifying manner, such as in song titles like "Diablo Negro" by "Gitanos" or discussions about "Zigeunerschnitzel" and "Mohrenkopf" (a type of pastry). The posts also touch upon political discourse, with mentions of "Nazis," "Antifa," "GRÜNE" (Greens), and "Linkspopulisten" (left-wing populists), often in the context of political arguments or accusations. The language used is frequently aggressive and polarising, reflecting deep societal divisions and prejudices.
Selected News items
• Préfet, sous-préfet, maire et ministre de l'Intérieur coordonnés pour évacuer les gens du voyage
• 'Incidenti naj se obravnavajo na podlagi dejanj, ne etnične pripadnosti' | 24ur.com
Sexism
Highest share of neutral messages overall (63.6%).
Medium- and high-toxicity content remains the lowest across all baselines (14.16% and 0.22%), indicating comparatively less polarised discussions.
Anti-Refugees
Highest share of neutral content among all categories (49.3%), followed closely by anti-Muslim narratives.
Medium-toxicity content is relatively low (19.1%), while high-toxicity remains limited (0.38%).
Anti-Muslim
Neutral share of 47.5%, second highest after sexist and anti-refugee content.
Medium-toxicity messages account for 23.7%.
High-toxicity content reaches 1.29%, the second highest across baselines.
Anti-Roma
Neutral share of 45.2%, but this baseline stands out for its volatility.
Medium-toxicity content is elevated (29.8%), one of the highest proportions.
Records the highest share of highly toxic messages (1.51%).
Anti-LGBTQ+
Lowest neutral share (31.4%), suggesting strongly polarised discourse.
Highest proportion of low-toxicity content (44.2%).
Medium- and high-toxicity shares remain lower than antisemitism or anti-Roma.
Antisemitism
Neutral share of 32.6%, also reflecting polarised discussions.
Medium-toxicity content makes up 31.1%.
High-toxicity content reaches 1.0%, the second highest after anti-Roma.
Hate speech by category
These data reveal clear patterns across the six monitored baselines. Antisemitic content is highly framed through racism (78.2%) and religion (45.1%). Political content is also highly prominent (40.1%) and posts containing threat content are the highest among the baselines (18.1%). Anti-Muslim discourse is dominated by religious framing (73.3%) and racism (45.6%), with politics (25.2%) and threats (16%) adding to its intersectional nature. Anti-LGBTQ+ content strongly overlaps with sexism (62.9%), reflecting its focus on gendered hostility. Sexism-related discourse is shaped by sexism itself (38%), but also intersects with politics (12.6%), ridicule (9.65%), and threats (9.84%), adding another layer of hostility. Anti-refugee narratives are primarily political (43.3%) and racist (37%), while threatening language is present in 13.53% of cases. Anti-Roma content is also heavily racialised (47.3%), with smaller but significant overlaps with politics (10.85%). Across all baselines, racism, politics, religion, and sexism emerge as the main vehicles of hate, with antisemitic and anti-Muslim narratives displaying the highest levels of violent threats.