Abstract A large body of research suggests that better-looking people are associated with various positive outcomes, an effect labeled the beauty premium. However, the vast majority of this evidence has been demonstrated in WEIRD (Western, Educated, Industrialized, Rich, and Democratic) countries, neglecting the role of cultural mechanisms in translating beauty into socio-economic outcomes. We close this gap by leveraging cultural differences embedded in an essential artifact of culture: language. Specifically, we establish a method for examining beauty associations in machine learning-based language models. Using this method, we replicate several standard findings in English. More importantly, we create a linguistic measure of the beauty premium and apply it to 68 languages. We provide new evidence that the beauty premium may be universal, with considerable heterogeneity across cultures. Furthermore, we discuss several new avenues for future research arising from our findings.