Abstract
<title>Abstract</title> <p>Online pornography is a major global media industry, yet its content remains structured by entrenched gender and racial stereotypes. Using Natural Language Processing (NLP), we analyze more than 250,000 video titles on the major platform Pornhub (2008–2024) to show how such stereotypes are embedded in the platform’s metadata. Gender and racial stereotypes emerge not only through the words applied to social categories but also through their grammatical and contextual framing. Titles overwhelmingly center on women, who are disproportionately described through appearance-, youth-, and smallness-based adjectives and positioned in voyeuristic, domestic, and transgressive settings. Racial stereotypes appear both as intensifiers and as specific reconfigurations of gendered scripts. Engagement analyses show that titles emphasizing large sizes, passivity, or transgressive locations attract the most attention, while racialized descriptors receive significantly lower engagement. Together, these findings demonstrate how linguistic structures in pornographic metadata reproduce and reinforce broader digital inequalities in gender and racial representation.</p>