Algorithms of arousal: an investigation of Pornhub videos and advertisements recommended to browser sessions without prior viewing

Abstract

This paper explores the characteristics of videos and advertisements recommended to browser sessions without prior viewing on Pornhub, focusing on the relationship between aggression and the race of female performers. To do so, it relies on a manual review of ads and on data obtained through the use of scripted browser instances visiting porn sites to model what recommendations are present in actual visiting users’ experience. The specific question it aims to answer is how the level and kind of aggression varies based on the female performer’s race/ethnicity in videos recommended to fresh browser sessions. The paper is inspired by recent writings by Adler (2024) and Srinivasan (2021) that draw attention to the harms the way porn today is consumed causes. Both discuss the view that sexual tastes are not self-determined and biologically essential but influenced by porn. As such, porn’s harm is conceptualized both in the direct and horrible toll its production can take but also in the way pornography that is widely consumed influences sexual norms, and as they are closely intertwined with sexuality, gender norms. The paper’s main findings are that in 2025, on Pornhub the ads shown to browser profiles without prior interaction are exclusively sexuality related, targeted towards men and mainly feature white women. For the videos recommended, based on the tags and categories found on the site, the ones that feature female performers of color also have aggressive terms more often. The effect is more pronounced for search results, but overall presence of aggressive tags or categories are higher for homepage recommendations. This finding is broadly in line with previous literature, most notably Shor and Golriz’s 2019 article Gender, Race, and Aggression in Mainstream Pornography, an important inspiration, with this paper having a significantly larger sample size but relying on external categorization for its operationalization of key categories.