Visual and Narrative Patterns of Online Misogyny: A Computer Vision Analysis of Telegram Chats

This article introduces a computer-vision pipeline designed to support feminist qualitative analysis of image-based hate speech against women in public Telegram channels. Using data collected between January and December 2024, the study responds to persistent challenges in feminist media research on online misogyny, particularly the scale of visual data, the multimodality of online interactions, and the limited accessibility of computational tools. We processed and clustered 2,122 images from a single public Telegram channel to identify recurring visual and narrative patterns associated with online misogyny. Rather than treating automation as a substitute for interpretation, clustering is mobilized as a dimensionality-reduction strategy that enables exploratory engagement with large visual corpora while preserving feminist commitments to contextualization, reflexivity, and power-sensitive analysis. The analysis reveals dominant patterns such as screenshots from other platforms, manipulated images, memes, cartoons, and gamified representations that sexualize, ridicule, and surveil women. These visual patterns operate within a broader multiplatform ecosystem in which women’s digital traces are captured, decontextualized, and re-presented as evidentiary material for misogynistic claims. The findings show that online misogyny is structured through intersectional systems of gender, race, age, and embodiment, contributing to the normalization of symbolic violence in low-moderation digital environments.