Since the late 2000s, the world has experienced an unprecedented platformization of cultural production. Cultural content is created, disseminated, and consumed on online platforms, leading to a state wherein algorithmic systems and coding infrastructure primarily mediate interaction with audiovisual media. The question this thesis addresses is how we can use existing algorithmic and software systems within platforms to study sexual desire and sexuality or, at the very least, the lust of their users. To do so, I assess the social norms unfolding within the content categorization and recommendation systems of Pornhub, a popular pornographic video online platform. My methodological approach is multi-modal since observational data (collected and found) are combined with interpretation and digital experiments. Conventional content analysis methods, such as term frequencies or semantic networks, are limited in their ability to capture the nuances of video category semantics. To work around this, I use a custom word embedding model treating categories as the model’s “vocabulary” and mapping them to a 100-dimensional semantic space in which similar words appear in proximity. Results showcase a promising and detailed semantic understanding of category meaning that can be further applied in studies of algorithmic bias of the platform’s video recommendations.