Data collection followed a so-called user-centred approach(Mangold & Stier, 2025). The starting point was a list of 136 public Telegram channels that had been identified by established experts in right-wing extremism research as clearly belonging to the right-wing extremist spectrum. Using Telegram’s open application programming interface (API), all messages from the 127 channels that were still active were retrieved and stored at two different points in time. The first data collection phase took place between August and September 2024. The second was conducted in the period surrounding the German federal election, from December 2024 to March 2025. The database of message texts also contains basic metadata, including channel IDs, user IDs, timestamps of publication, and interaction metrics (e.g. forwards). In a second step, all links contained in the messages were extracted, and information was recorded on how frequently each link appeared across the dataset. In a third step, only links to YouTube videos were retained. Metadata for these videos, comments on the videos, and video transcripts were then retrieved via the YouTube API. In a fourth step, a subset of messages was manually annotated. For this purpose, a codebook with detailed annotation guidelines was developed. Each message was assigned to one or more categories of hate speech.

The resulting data corpora will be made available for scientific reuse in the near future. Detailed information on access conditions and application procedures will be provided here.