Gender Bias in YouTube Exposure: Allocative and Structural Inequalities in Political Information Environments

A new study published on arXiv (April 2026) by Jipeng Tan and colleagues at Chinese universities reveals that YouTube's recommendation algorithm exhibits gender bias in political content exposure. Using a controlled social-bot field experiment, researchers created male-coded and female-coded profiles and tracked their recommendation trajectories. They found two types of bias: allocative (unequal distribution of issues, ideological leanings, and political entities) and structural (distinct clustering of recommended content). For example, female-coded profiles were more likely to receive content on social welfare issues, while male-coded profiles saw more economic and security topics.

Time-series analysis showed that these exposure pathways evolve over time, shaped by both community structures and individual profile dynamics. The researchers also built a simple collaborative-filtering model that reproduced the observed biases, confirming that the disparities stem from algorithm design rather than user behavior alone. The paper argues that such biases not only distort individual information environments but also reinforce societal gender inequalities. The findings highlight the need for algorithmic fairness audits and suggest that platforms like YouTube must consider both allocative and structural dimensions of bias in their ranking systems.