All Reports

Researchers from Open University and Politecnico di Torino find female-dominated occupations face higher exposure to Large Language Models

Bot Mutiny |

Summary

A gendered analysis of labor market data shows that while AI exposure for men is concentrated in high-wage roles, women in low-skilled and low-paid positions face significant exposure to task automation and job restructuring.

A paper posted to arXiv on September 18, 2026 examines how exposure to AI varies across male- and female-dominated occupations. It is titled "When AI Enters the Workplace, Who Faces Greater Risks? A Gendered Analysis." The authors are affiliated with the Centre for Protecting Women Online at the Open University (UK), Politecnico di Torino (Italy), Nokia Bell Labs (UK) and the Fawcett Society (UK). The arXiv listing is a version 1 posting in the cs.CY category. The text read does not say whether the paper has been peer reviewed.

Method

The authors built a dataset linking O*NET occupational data with U.S. Census data on gender-disaggregated income distributions. They combine two indices. The Anthropic Index captures exposure to Large Language Models (LLMs) associated with recent advances in generative AI. The Artificial Intelligence Index (AII) captures exposure to broader AI innovation, such as robotics, based on patent data.

Exposure by Pay Level

The authors find that AI exposure in male-dominated occupations is generally concentrated in higher-skilled and higher-paid occupations. Female-dominated occupations show relatively uniform levels of exposure across both high-skilled, high-paid and low-skilled, low-paid occupations. The paper says this relationship does not hold within female-dominated occupations.

LLMs and Broader AI

The authors find that LLM-related exposure is higher in female-dominated occupations, while exposure to broader AI innovation remains more concentrated in male-dominated occupations.

They add a caution: AI exposure does not necessarily imply realised automation or displacement effects. Exposure measures capture the extent to which occupational tasks are likely to be affected by AI systems, whether through augmentation, restructuring or automation.

What the Authors Infer

The paper then triangulates its results with existing literature. Its authors say this suggests men are more likely to benefit from AI as a complement to high-skilled work. They say women in lower-skilled and lower-paid, female-dominated occupations are more exposed to forms of AI associated with task automation and job restructuring. In those occupations, they write, exposure is "less likely to translate into productivity-enhancing augmentation." They say it is more likely to involve reorganisation or substitution of tasks, increasing the risk of reduced wages, limited career progression and job displacement. This is an interpretation drawn from prior literature, not a result measured in the paper's own data.

Adoption Gap

The paper cites Aldasoro et al. (2024), a U.S. survey study, which finds that men report substantially higher use of generative AI tools than women, 50% versus 37%. That study identifies differences in knowledge about the technology as the primary driver, accounting for roughly three-quarters of the difference. The remainder is explained by gender differences in trust and privacy concerns.

The paper also cites Otis et al. (2024), which warns that persistent differences in adoption may create a self-reinforcing cycle. Lower participation of women in AI use may lead to systems trained on data that insufficiently reflect women's needs and preferences. The authors add that participation in adult learning and training is substantially lower among low-skilled workers, citing OECD 2018.

Updated October 2, 2026 after a source verification pass.