Quarterly Insights: Online Hate and Toxicity Trends (Q1 2025 Report)

This report is the second in a series providing a quarterly analysis of online harm trends. This edition examines patterns observed in January, February, and March 2025.

Data for March, February, and March was based on 11,7 M messages in 24 languages, across 6 social media platforms: Reddit, X, 4chan, Gab, YouTube, and Facebook.

Toxicity over time 

This timeline illustrates the average toxicity level on social media in Q4 2024.

The toxicity score averaged at 0.21 with a slight increase at the end of January but without major fluctuations.

Are you unsure about the terminology we use in this report? Visit our FAQ for more details.

Regional breakdown

This section underlines toxicity trends across four different European regions*: Western Europe, Southern Europe, Northern Europe, and Eastern Europe. 

From January to March 2025, the toxicity score displayed distinct regional patterns across Europe. Western Europe recorded the highest toxicity, with an average of 0.22, peaking at 0.26 at the end of January. Southern Europe followed closely, maintaining a relatively stable score around 0.20–0.22 throughout the period. Eastern Europe showed steady toxicity levels, fluctuating between 0.11 and 0.16. In contrast, Northern Europe reported the lowest toxicity scores, beginning at 0.09, dipping to 0.06 early in January, and gradually rising to 0.15 by late March. Overall, the data indicates that toxicity remained most pronounced in Western and Southern Europe.

*Our regional distinction in Europe is based on the languages monitored via the dashboard, grouping countries accordingly to reflect linguistic and geographical patterns. The monitored languages include Danish, Estonian, Finnish, Lithuanian, and Swedish (Northern Europe); Croatian, Czech, Hungarian, Polish, Slovak, Slovenian, Bulgarian, Romanian, Russian, Macedonian (Central and Eastern Europe); French, German, English, Dutch (Western Europe); and Spanish, Italian, Portuguese, Greek and Arabic (Southern Europe).

VLOPs vs non-VLOPs

The following graph compares the average toxicity levels on Very Large Online Platforms (VLOPs)—including Facebook, X, and YouTube—versus non-VLOPs such as 4chan, Gab, and Reddit, between January and March 2025.

4chan exhibited the highest toxicity overall, averaging around 0.38, reinforcing its position as the most toxic platform in the dataset. Reddit recorded the second-highest average toxicity, with levels consistently ranging between 0.30 and 0.33 in February and March. This places it above fringe platform Gab (average 0.29) and significantly higher than mainstream platforms like Twitter (0.21–0.22) and YouTube (0.24–0.27).

Reddit also experienced a sharp spike in toxicity during the first half of January, jumping from 0.18 to 0.30 in a single day. While Facebook generally maintained the lowest toxicity levels (around 0.12–0.15), it registered a sudden and significant spike** in the last week of January, increasing from 0.12 to 0.27 on January 19 -a 0.13- point rise in one day.

**This sharp increase may be partially attributed to an outage in data collection during that period, which could have skewed the results.

These findings highlight how mainstream platforms like Reddit and Facebook experienced rapid spikes in toxic content, with Reddit standing out for its persistently elevated toxicity levels, second only to 4chan.

Hate speech by category

This table illustrates the prevalence of different hate speech categories across various targeted groups.

Sexism-related content shows the highest incidence of sexist language (49%) alongside political narratives (36.04%). Anti-LGBTQ+ hate stands out for its high sexism (97.72%), obscenity (32.92%), and profanity (17.37%), reflecting aggressive and gendered hostility.

Hate speech targeting Muslim communities is overwhelmingly racialised (75% racism) and religiously charged (157%), while antisemitic messages are also heavily racialised (151% racism) and religiously framed (103%), with a significant political dimension (97%). Anti-refugee discourse is highly politicised (103%) and racialised (61%), while anti-Roma narratives show marked racism (82%) and frequent use of ridicule (11%) and threatening language (15%). These findings highlight how hate speech is tailored through specific framings, often combining racism, sexism, and political discourse depending on the targeted group.

Semantic analysis per platform

This table outlines the intensity and framing of hate speech across six online platforms: Facebook, Reddit, Twitter (X), Youtube, 4Chan, and Gab. Reddit stands out with the highest overall levels, particularly in obscenity (117.67%), sexism (100.26%), religion-related hate (89.97%), and political content (150%). It also exhibits a strikingly high presence of ridicule (61.10%) and threats (127.15%), indicating a complex and multifaceted expression of toxicity.

YouTube also shows notable toxicity, especially in religion-related speech (265.65%), politics (216.36%), and racism (122.54%), suggesting a highly ideological and radicalised use of the platform. Gab, known for its lack of content moderation, presents elevated levels of political hate (268.87%), religion (55.92%), and threats (83.86%).

4chan maintains high scores across almost all categories, with particularly high levels of racism (96.58%), sexism (52.16%), and ridicule (35.08%), affirming its reputation as a hub for unmoderated toxic speech. In contrast, Facebook and Twitter show comparatively lower levels across most categories, although Twitter still registers a notable presence in sexism (18.26%), politics (52.51%), and racism (32.09%), while Facebook displays modest but present levels of political toxicity (24.10%). These differences reveal how each platform facilitates distinct forms and intensities of hate speech.