The AI Gender Gap

There is a conversation happening right now about artificial intelligence — about jobs, creativity, productivity, and the future of work. It is loud, it is everywhere, and it is mostly being had without women.

Not because women aren't listening, but because the people building the technology, and the systems deciding whose voices get included in its training data, have a serious problem that isn't being talked about nearly enough.

This article is about what it means when the most powerful technology in human history is designed, built, and deployed by an overwhelmingly male workforce — and why, if you care about equity, the future of STEM, and the world AI is going to create, this is the conversation you need to be having.

First, the numbers

The AI industry has a gender problem that mirrors, and in some ways exceeds, the broader STEM gender gap.

• Less than 30% of the global AI workforce is female. (World Economic Forum, Global Gender Gap Report 2023)

• 15% of senior AI roles are held by women — half the already-low overall representation. (Indiana University Centre for Women & Technology, 2025)

• 44% of AI systems studied showed measurable gender bias. (Berkeley Haas Centre for Equity, Gender and Leadership, 2024)

• Around 29% of female-dominated occupations face high exposure to AI automation, compared with 16% of male-dominated ones. (International Labour Organisation, March 2026)

Read that last one again. The jobs most at risk of being automated away by AI are disproportionately women's jobs, and the people building that AI are disproportionately men. This is not a coincidence — it's a design failure.

What bias in AI actually looks like

It's easy to think of AI bias as an abstract, technical problem. It isn't. It shows up in specific, concrete ways that affect real women's lives.

In hiring

In 2018, Amazon scrapped an AI recruiting tool after discovering it was systematically downgrading CVs that included the word "women's," as in "women's chess club" or "women's college." The tool had been trained on historical hiring data reflecting decades of male-dominated hiring decisions — it learned from the past and projected it into the future. By 2024, UNESCO research confirmed that gender bias in AI hiring tools continues to penalise women by reproducing the same regressive stereotypes. (Reuters, 2018; UNESCO, 2024)

In language models

A 2024 UNESCO study found that large language models like ChatGPT regularly associate women with terms like "family," "home," and "children," and men with "executive," "business," "career," and "salary." One AI model associated women with domestic roles up to four times more often than men. These aren't edge cases — they're the default. (UNESCO, Bias Against Women and Girls in Large Language Models, 2024)

In images

A 2024 study published in JAMA Open Network found that when asked to generate images of physicians, AI text-to-image tools defaulted to depicting white men. When a Turkish artist prompted an AI to write a story about a doctor and a nurse, the AI made the doctor a man and the nurse a woman every single time, regardless of how she rephrased the prompt. (JAMA Open Network, 2024; UN Women, 2025)

In the workplace

Research from Harvard Business Review found that female engineers who used AI for code generation faced a 13% competence penalty — more than twice the 6% penalty applied to men for the same AI-assisted work. The work was identical. The penalty wasn't. (Harvard Business Review, via Indiana University Centre for Women & Technology, 2025)

AI mirrors the biases of the world it was trained on, and that world was not designed with women in mind.

Why this is a STEM pipeline problem

AI bias isn't just a consequence of the gender gap in STEM — it actively makes the gender gap worse. When a 13-year-old girl searches for images of "scientist" and sees mostly white men, when a young woman's CV is filtered out by a biased algorithm before a human ever sees it, when voice assistants respond to harassment with compliance rather than pushback, these systems send a message about who belongs and who doesn't.

And the people who could fix this are the same people who have been told, since childhood, that this field isn't for them.

This is the loop: fewer women in STEM means fewer women building AI. Fewer women building AI means more biased AI systems. More biased AI systems mean more signals to young women that they don't belong — which means fewer women in STEM.

Breaking that loop requires intervention at every stage: girls encouraged into STEM early, mentorship, scholarships, community, organisations like HerOrigin making the pipeline visible and accessible, and an industry willing to treat this not as a diversity initiative but as a fundamental quality problem.

What good looks like

The news isn't all bleak. There are examples of what happens when women are included in AI design, and they're worth knowing about.

• UN Women's AI Casebook on Gender and Agriculture showcases 26 scalable AI solutions designed specifically for female farmers, improving crop planning, financial access, and climate resilience. When women were included in the design process, the technology worked better — for everyone. (UN Women, 2025)

• Deloitte research found the proportion of US women experimenting with and using generative AI tripled between 2023 and 2024, outpacing the growth rate for men. (Deloitte Connected Consumer Survey, 2024)

• Argentina and Malaysia have already achieved gender parity among R&D researchers, with over 53% of their research workforce identifying as women. (UNESCO Institute for Statistics, 2024)

What this means for HerOrigin

At HerOrigin, we tell stories. We connect young women with role models, mentors, and each other. We believe visibility changes what feels possible, and that a girl who sees herself reflected in STEM is more likely to stay in it.

The AI gender gap makes this work more urgent, not less. The stakes of women's absence from STEM are no longer just about career equity — they're about who gets to shape the intelligence that will run hospitals, hire workers, write laws, teach children, and diagnose disease.

If that intelligence is built without women, trained on data that reflects a world where women are secondary, it will build a future that reflects the same.

We can't afford that. And we don't have to accept it. The future of AI is not inevitable — it's a design choice, and we need women at the table when those choices are made.