Monthly Report May 2025

Data for May was based on 2,594,657 M messages in 11 languages, across 6 social media platforms: Reddit, X, 4chan, Gab, YouTube, and TikTok.

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

Average toxicity

In May, the overall average toxicity level was 0.21. This figure showed moderate fluctuations throughout the month, with notable peaks occurring during the second and fourth weeks before stabilising at 0.22 toward month's end.

Baseline channel analysis

This section provides an overview of the average toxicity levels among the baselines over May. Antisemitism maintains the highest average toxicity score at 0.34, demonstrating consistent elevation throughout the monitoring period. Anti-Roma content showed the second-highest average at 0.30, representing sustained targeting of this community. Anti-LGBTQ+ and Anti-Muslim narratives both registered significant toxicity levels at 0.29 and 0.28, respectively, while anti-refugee content maintained a moderate level at 0.23. Despite recording the lowest average toxicity score at 0.19, sexist content accounted for the highest volume of toxic posts, with 288,371 instances - over three times the number of anti-Roma posts and more than double those targeting LGBTQ+ individuals.

Baseline Sexism

Volume: 288,371 posts identified as toxic content.

Platforms: Most posts (70%) originated from X, with TikTok (13%) and Reddit (11%) also contributing significantly.

Toxicity: The average toxicity score was 0.19. Posts with extreme toxicity (0.8 or higher) totaled 18,109, predominantly on 4chan.

Themes:

• Sexism was the most prevalent category (50%), followed by Violence (13%).

• Common toxic keywords included bitch (English), slut (English), whore (English), حد (Arabic; "limit" or potentially a reference in religious/legal context), and freak (English).

Key Insight: Despite the high volume of posts, sexist content maintains relatively lower toxicity scores, though extreme toxicity remains concentrated on 4chan with misogynistic language dominating the most harmful discourse.

Baseline Anti-LGBTQ+

Volume: 83,711 posts identified as toxic content.

Platforms: Most posts (81%) originated from X, with Reddit (11%) and TikTok (4%) also contributing significantly.

Toxicity: The average toxicity score was 0.29. Posts with extreme toxicity (0.8 or higher) totaled 4,692, predominantly on Gab.

Themes:

• Sexism was the most prevalent category (76%), followed by Violence (10%).

• Common toxic keywords included شاذ (Arabic; "pervert" or "deviant"), slut (English), bitch (English), whore (English), and sluts (English).

Key Insight: While X hosts the majority of content, Gab concentrates the most extreme toxicity, with anti-LGBTQ+ discourse heavily intersecting with misogynistic language and slurs.

Baseline Anti-Muslim

Volume: 128,589 posts identified as toxic content.

Platforms: Most posts (87%) originated from X, with TikTok (5%) and Reddit (4%) also contributing significantly.

Toxicity: The average toxicity score was 0.28. Posts with extreme toxicity (0.8 or higher) totaled 10,419, predominantly on 4chan.

Themes:

• Religion was the most prevalent category (76%), followed by Violence (19%).

• Common toxic keywords included moro (Filipino/Spanish; pejorative for Muslim groups), moros (Spanish; derogatory term for Muslims), mora (Italian/Spanish slang; can carry xenophobic connotations), المشركين (Arabic; "polytheists," often used in a hostile religious context), and حد (Arabic; "limit" or potentially a reference in religious/legal context).

Key Insight: Religious hate speech dominates anti-Muslim content, with 4chan hosting the highest concentration of extremely toxic posts that blend religious hostility with violent rhetoric.

Baseline Anti-Refugee/Migrants

Volume: 36,076 posts identified as toxic content.

Platforms: Most posts originated from X (85%), with Reddit (5%) and TikTok (4%) also contributing significantly.

Toxicity: The average toxicity score was 0.23. Posts with extreme toxicity (0.8 or higher) totaled 1,350, predominantly on 4chan.

Themes:

• Politics was the most prevalent category (65%), followed by Violence (16%).

• Common toxic keywords included fachkräfte (German; sarcastic use meaning "skilled workers," often referring to migrants), white genocide (English; conspiracy theory terminology), negro (Spanish/Italian; racial slur), حد (Arabic; "limit" or potentially a reference in religious/legal context), and "migrants" (English; in scare quotes to signal hostility).

Key Insight: Political discourse dominates anti-refugee narratives, with extreme toxicity concentrated on 4chan and characterised by conspiracy theories and racial slurs targeting immigrant populations.

Baseline Antisemitism

Volume: 153,099 posts identified as toxic content.

Platforms: Most posts (88%) originated from X, with Reddit (4%) and 4chan (2%) also contributing significantly.

Toxicity: The average toxicity score was 0.34. Posts with extreme toxicity (0.8 or higher) totaled 9,798, predominantly on 4chan.

Themes:

• Racism was the most prevalent category (76%), followed by Violence (23%).

• Common toxic keywords included الصهاينة (Arabic; "Zionists"), kill jews (English), niggers (English slur), kikes (English antisemitic slur), and قتل (Arabic; "kill").

Key Insight: Antisemitic content maintains the highest toxicity levels across all baselines, with 4chan concentrating the most extreme racist and violent rhetoric despite X hosting the majority of posts.

Baseline Roma

Volume: 21,555 posts identified as toxic content.

Platforms: Most posts originated from X (82%), with Reddit (7%) and TikTok (4%) also contributing significantly.

Toxicity: The average toxicity score was 0.30. Posts with extreme toxicity (0.8 or higher) totaled 1,502, predominantly on 4chan.

Themes:

• Racism was the most prevalent category (86%), followed by Violence (8%).

• Common toxic keywords included gitanos (Spanish; "gypsies," often used pejoratively), gitano (Spanish; singular form), slut (English), os ciganos (Portuguese; "the gypsies," often used negatively), and sluts (English).

Key Insight: Anti-Roma content demonstrates exceptionally high levels of racist discourse, with 4chan hosting the most extreme toxicity and intersecting anti-Roma sentiment with misogynistic language.

This graph illustrates the distribution of toxicity levels—Neutral, Low, Medium, and High—across our monitored baselines. Anti-Roma and anti-Muslim datasets had the highest percentage of highly toxic messages, with a proportion of 2.84% and 2.64% respectively, while the antisemitism baseline shows the lowest proportion of neutral messages (26.02%), denoting particularly polarised and/or toxic discussions.

Hate speech by category

The hate speech categorization data reveals distinct patterns across the six monitored baselines: antisemitic content is dominated by racism (74.57%), while more than 1 in 5 messages contains language related to threat (22.71%). Anti-Muslim discourse is primarily religious (67.89%) with significant racist elements (42.77%). Anti-LGBTQ+ content heavily intersects with sexism (57.64%), and anti-refugee narratives are predominantly political (48.19%) and racist (39.87%). Anti-Roma content is primarily racist (55.10%), while general discussions around gender are mostly associated with politics (16.89%), apart from the expected sexism (31.76%). Across all categories, racism, politics, religion, and sexism emerge as the primary rhetorical vehicles for targeting different communities, with antisemitic and anti-Muslim content showing the highest levels of violent threats, indicating that these communities face particularly severe forms of online hate speech.