Bot comments Report September 2025
The dataset of toxic bot comments detected in September 2025, spanning from September 01 to September 24, included over over 3.25K English messages (n=3,281). The average toxicity score is 0.20, which is quite low. However, there is a notable peak (2 × SD) on September 08, 2025 (see the list of events that could have influenced it). About 35 messages (1%) are very toxic, with a score > 0.8.
Toxicity often involved Politics (15%), Religion, Racism and/or Contempt.
Threatening content: about 10% of messages was categorized as Threatening.
Disinformation: about 1% of the content was categorized as Disinformation.
The most frequent combination: Threatening + Ridicule.
Some common toxic keywords include: slayer, rape, demon, jewish donors, flesh, final solution, propaganda, zio, murder, and slave.
The word cloud below highlights the 50 most frequently used toxic keywords in August 2025. The size of each word reflects its frequency relative to other terms. Notably, gun was the most frequently used keyword, followed by allah, killed and oppressed.
The table below highlights the different toxic categories detected within the September 2025 dataset, along with their respective frequencies (represented by black dots and further explained below). In this month, the category Jews contained bot comments with the highest toxicity score: 0.70. The categories Politics, Religion and Threat were most prominent with toxic bot comments content.
The barchart below shows that most of the content fell within the Religion category. Content with a toxicity score > 0.8 appeared in all categories (see the explanation of the different categories here).
The keyword network below illustrates the main focus of toxic bot comments detected in August 2025. Different colors represent clusters of words frequently used together, revealing the bot’s primary narratives. In this case, one can see two different narratives with many keywords: one surrounding the word die and one surrounding the word Israel.