Quarterly Insights: Online Hate and Toxicity Trends (Q4 2025 Report)

This report is the fifth and last in a series providing a quarterly analysis of online harm trends. This edition examines patterns observed in October, November, and December 2025. 

Data for October, November, and December was based on 62.06 million messages in 27 languages across 20 social media platforms, among which were Reddit, X, 4chan, Gab, YouTube, Facebook, Thread, TikTok, and Instagram.

Toxicity over time 

This timeline shows the average toxicity levels on social media during Q4 2025.

In October, November, and December, the average social media toxicity had minimal fluctuation, ranging between 0.19 and 0.22, with a small increase from 0.20 to 0.22 mid-December. 

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. 

Western Europe consistently recorded the highest toxicity levels, hovering around 0.21–0.22 throughout October and November before showing a gradual increase in mid-December, peaking at 0.23 by the end of the year. These trends are based on an analysis of over 45.4 million social media posts, providing a robust picture of toxicity patterns in the region. Southern Europe followed a similar trajectory at slightly lower levels, generally ranging between 0.20 and 0.22, with a rise toward late December that mirrored trends in Western Europe, based on approximately 7.3 million posts. Northern Europe, drawing on around 536,000 posts, mostly fluctuated between 0.09 and 0.11, with a brief mid-December uptick reaching an average of 0.14. Eastern Europe displayed the lowest toxicity levels overall, averaging around 0.09–0.11 across the period, with only minor increases toward the end of the year, based on roughly 4.1 million posts. Overall, regional disparities persisted, with Western and Southern Europe showing higher and slightly increasing toxicity levels, while Northern and Eastern Europe remained comparatively low and stable.

VLOPs vs non-VLOPs

The following graph compares the average toxicity levels on Very Large Online Platforms (VLOPs) - including Facebook, X, YouTube, Instagram, and TikTok - with those on non-VLOPs such as 4chan, Gab, Reddit, and Thread, between October and December 2025.

Among non-VLOPs, 4chan, WordPress, and Gab continued to register the highest toxicity levels overall. 4chan remained the most toxic platform, consistently averaging between 0.40 and 0.45. WordPress followed closely, maintaining elevated toxicity levels generally between 0.27 and 0.40, with multiple spikes above 0.40 toward the end of November and December. Gab showed slightly lower but still high toxicity, ranging between 0.27 and 0.33. 9gag occupied a mid-range position, averaging around 0.24-0.27, indicating moderate but persistent toxicity across the period. Telegram remained comparatively less toxic among non-VLOPs, with stable values around 0.13–0.16, showing only mild increases in mid-December.

Among VLOPs, toxicity levels were lower and more stable. Twitter (X) maintained moderate but consistent levels around 0.22–0.25, with slight increases toward mid-December. YouTube showed relatively low and steady toxicity, generally between 0.16 and 0.18, with a small upward shift in December. Facebook and Instagram continued to register low toxicity overall, typically ranging from 0.12 to 0.16, with only minor short-lived fluctuations. TikTok consistently displayed the lowest toxicity levels of all monitored platforms, remaining around 0.10–0.13 throughout the period. Threads showed stable, low-to-moderate toxicity, averaging roughly 0.18–0.22, with a gradual increase toward late December.

Overall, non-VLOPs continued to exhibit significantly higher and more volatile toxicity levels than VLOPs, with 4chan and WordPress standing out as persistently higher toxic environments. In contrast, platforms covered by the DSA’s VLOP designation demonstrated more consistent moderation patterns and generally lower, more stable toxicity averages over the period.

Hate speech by category

The percentage distribution highlights distinct and highly concentrated patterns of toxicity across baselines, with a small number of categories dominating each narrative. Antisemitic content is overwhelmingly driven by racism (82.68%), alongside substantial intersections with religion (41.27%) and politics (37.96%), underscoring the deeply ideological and conspiratorial nature of this discourse. Anti-Muslim narratives show a similarly pronounced concentration, with religion appearing in nearly 80% of posts and racism in just over 49%, reinforced by a strong political dimension (27.61%) and elevated levels of threats (21.61%). In anti-LGBTQ+ content, sexism is the defining feature, accounting for 62.77% of posts, exceeding all other categories, while ridicule (11.55%) and obscenity (9.10%) further shape a degrading tone. Anti-refugee discourse is primarily politicised (45.86%) and racialised (38.17%), with threats also prominent (16.54%), reflecting securitised and exclusionary framing. Anti-Roma content, though lower in intensity overall, remains predominantly racialised (28.34%), with politics (11.17%) and threats (8.84%) as secondary elements. Finally, the sexism baseline itself is largely characterised by explicit sexist language (36.67%)