Most conversations about women and AI focus on who is using it. The more revealing question is who is building it. The people who design these tools decide, quietly, what the tools are good at, whose problems they solve, and whose they overlook. On that measure, women in AI are still a small minority, and the gap sits much deeper than adoption.
The numbers are blunt. Women make up about 22% of AI professionals worldwide, roughly 66,000 out of an estimated 300,000 specialists, according to World Economic Forum figures cited across the field. The share shrinks as the work gets more senior and more technical. Women hold under 14% of senior executive roles in AI, and only around 12% of AI researchers are women. Computer Weekly, reporting on the same talent data, put it plainly: less than a quarter of the people shaping this technology are women.
Where women in AI actually stand
It helps to separate two different gaps. One is the adoption gap, the fact that women often reach for AI tools later or more cautiously than men. That gap is real, and in some places it is closing fast. The other is the creation gap, the fact that women are largely absent from the rooms where the models, the datasets and the product decisions are made. The first gap affects who benefits. The second shapes what actually gets built.
The second gap is the one that compounds. Interface, a European research organization, has argued that a narrow talent pool feeds directly into narrow products: fewer perspectives on fairness, on safety, on the everyday situations a tool needs to handle well. When most builders share a background, the blind spots get shipped along with the features.
Why this is structural, not personal
It would be easy, and wrong, to read these numbers as a confidence problem. The barriers are structural. Technical AI roles sit at the end of a long pipeline, and women are filtered out at several points along it: who gets encouraged into computer science, who gets hired onto research teams, who gets the funding to start an AI company, who gets promoted into the senior roles where the real decisions are made. Each step is a gate. Each gate leaks.
Trust plays a role too, and it is not irrational. Deloitte’s research on women and generative AI found women express less confidence than men that AI providers will keep their data secure. When you have less reason to trust how a system was built, and you had no hand in building it, caution is a reasonable response, not a deficit. Cherie Blair has said women are “not foolish” to be wary of AI, and that the honest fix is to make the tools, and the industry behind them, more worth trusting.
There is a hopeful line in the data worth holding onto. By 2026, an estimated 35% of AI research papers are projected to have a woman as lead author, up meaningfully from a few years ago. Progress at the research edge does not close the executive gap on its own. But it is a sign the pipeline is not fixed in place.
One thing that moves the needle
For a woman running a small business or weighing a move into the field, the useful step is not to wait until the industry fixes itself. It is to move one rung closer to building, not just using. That can be small. Learning to shape a tool rather than only prompt it, joining one community where women are doing technical AI work, or mentoring someone a step behind. None of that closes a 22% gap alone. But representation in AI has always been built by people who decided to take up more space than they were handed.
The broader pattern connects to ground covered here before, on the gap between adopting AI and gaining from it, the funding gap facing women founders, and the recognition women lose for the same work. Who uses AI, who profits from it, and who builds it are three versions of one question. The building one may matter most, because it decides all the others.
So the question worth sitting with is not whether more women will use AI. They already are. It is whether the next generation of these tools will be shaped by the people they are meant to serve.
