Pornography-related discourse on X (Formerly Twitter): a topic-modeling analysis of more than half a million posts across four languages (2014–2024)

Pornography use is widespread, yet there is limited evidence on how pornography is discussed and framed in everyday online environments, particularly from social-media-listening approaches that analyse large volumes of spontaneous public discourse. Social-media listening may complement self-report surveys, which are vulnerable to social-desirability biases, by providing access to large volumes of spontaneous public discourse. This study used machine-learning-based topic-modeling methods to describe pornography-related Xeets on X (formerly Twitter) between 2014 and 2024 in four languages (English, Spanish, Catalan and Basque). Public Xeets in the four target languages were retrieved using the Tweet Binder search engine based on general query terms related to pornography and sexually explicit content. Eligible Xeets were publicly accessible, contained at least one predefined query term, were published between 2014 and 2024, and were written in one of the target languages. Language-specific topic models were estimated using BERTopic, and the most prevalent topics were qualitatively labeled by a multidisciplinary team. A total of 634,746 Xeets were analyzed. English-language Xeets were dominated by topics related to pornography-related scandals and pornography involving young individuals, whereas Spanish Xeets mainly referred to pornography consumption and crimes. Catalan Xeets frequently addressed violent or illegal pornography and Basque Xeets largely contained explicit content alongside a smaller share of social, legal and educational debates. Temporal patterns differed across languages and reported increases in pornography consumption during the COVID-19 pandemic did not correspond to clear increases in pornography-related Xeets. This large-scale, multilingual analysis identifies language-specific topics and temporal patterns in pornography-related discourse on X without directly measuring underlying attitudes, perceptions or sentiments. The descriptive findings may generate hypotheses for future public-health communication research, while highlighting the need for additional audience research, qualitative analysis, and sentiment-focused studies before the findings may inform awareness campaigns or culturally tailored messaging.