The AI boom in women’s health could reinforce old biases

As AI attracts growing attention in women’s health, researchers warn that reductive categories and expanding data surveillance could deepen the inequalities technology promises to address.

Perspective: Beyond the hype of AI as a panacea for women’s health. Image Credit: Faizal Ramli / Shutterstock

A recent perspective article in the journal npj Women’s Health critically examined the growing enthusiasm for artificial intelligence (AI) in women’s health research and practice. The authors highlighted the risks of treating AI as a cure-all and emphasized the importance of addressing structural and social factors to achieve gender equity in health outcomes. 

Rising Expectations for Artificial Intelligence

Gender and sex inequities continue to shape medical research and healthcare delivery. Women, as well as transgender and gender-expansive people, often receive poorer care and experience poorer health outcomes. Recent US executive orders have further limited research into gender, sex, and health disparities, likely exacerbating these problems.

These restrictions impose a simplistic, binary view of gender and sex and restrict research into their complexities. At the same time, there is growing optimism about AI’s potential to address these barriers. AI is frequently described as a means to generate personalized health insights, build tailored disease models, and gather individualized data.

However, framing AI as a universal fix for women’s health risks overshadows deeper systemic issues. Key concerns include the persistence of reductive gender and sex categories in research, the risk that uncritical adoption of AI will encode these flawed assumptions, and the possibility that such trends may perpetuate or worsen existing inequities in healthcare. Addressing these challenges demands more than attention to privacy or bias; it requires engagement with the underlying complexities of gender, sex, and technology.

Limitations of Binary Gender Frameworks

Much mainstream women’s health research continues to rely on narrow, binary definitions of gender and sex, ignoring the complex realities shaped by interactions among biological, social, and cultural factors, including race and class. This limits understanding and fails to include diverse identities, while transgender, intersex, and gender-diverse people already face disproportionate exposure to discriminatory medical practices and inadequate care.

Policies and research practices often default to binary categories, as seen in the National Institutes of Health (NIH) Sex as a Biological Variable (SABV) policy, which requires animal research to include both female and male animals and report findings by binary sex category. This approach sidelines broader gender considerations and may offer limited insight into individual health outcomes, even within cisgender populations. The paper contends that AI technologies frequently encode these limitations, perpetuating outdated assumptions rather than advancing a more nuanced understanding.

The Pitfalls of Uncritical AI Adoption

AI is frequently promoted as a powerful solution in medical research, but its definition remains vague, and its capabilities are often overstated. The paper cautions that this uncritical acceptance can reduce accountability and prioritize predictive accuracy over causal or mechanistic understanding. Machine learning is commonly used to identify differences, particularly in sex-based research, even when models rely on simplistic or flawed assumptions.

Medical research often defaults to binary sex categories without questioning their relevance or origins. Some attempts to reduce bias in AI can entrench these divisions, and the belief in AI’s objectivity can turn weak findings into claims about inherent differences. As a result, AI models can reproduce and amplify existing social inequalities while obscuring the limitations of binary thinking, hindering progress toward equitable healthcare. Although these problems predate AI, the authors contend that current AI hype may further normalize and legitimize them.

Data, Surveillance, and Deepening Inequities

Framing AI as a cure-all for women’s health can deepen inequities. The ongoing drive for more data often distracts from addressing flawed research questions and systemic barriers. Private actors can access and monetize intimate health and behavioral data, and the focus on technological innovation frequently shifts attention from needed structural change.

According to the authors, many AI-powered apps that promise empowerment depend on extensive surveillance and data extraction, raising privacy concerns and shifting responsibility for health onto individuals. This self-surveillance can undermine psychological well-being and depoliticize health disparities. In clinical settings, prioritizing data over lived experiences can silence patients.

Globally, the paper argues that many AI health tools extract data from vulnerable populations in the Global South under “AI for social good” narratives, but may distract from real public health needs. These trends sometimes replace essential care with technology and may fail to improve outcomes, as AI hype accelerates their adoption and expands their reach.

Rethinking Solutions for Gender Equity in Health

The authors write that research funding for women’s health and gender-equitable healthcare has declined dramatically, increasing interest in AI solutions. However, equity needs careful evaluation at every stage of the AI pipeline, from problem definition to data collection and use. Technology alone cannot resolve health inequities, and social and structural changes remain essential.

In their assessment, many AI applications in women’s health preserve binary categories and expand surveillance, often hiding deeper structural causes of disparity. Addressing bias, transparency, and accountability in AI requires scrutiny of the categories and assumptions behind these systems.

The authors recommend that researchers focus on relevant social and biological factors, involve affected communities, and weigh the risks of misrepresentation and surveillance against the need for representative data. The emphasis should move from who is included in datasets to how and why inclusion happens. The potential of AI depends on a nuanced understanding of sex, gender, and context, but meaningful progress in health equity requires a broader commitment to healthcare for all.

The perspective was published online as an unedited Article in Press and will undergo further editing.