AI-generated article. This article was researched and drafted using AI tools and published automatically, and its featured image was generated by AI. Facts are drawn from the sources cited in the text.
Two job openings sit side by side. Same company, same seniority, similar hours. One has the words “AI” or “machine learning” somewhere in the requirements and pays roughly double. Guess which one a woman is statistically far less likely to be hired into. That imbalance now has a number attached, and it is the clearest picture yet of the AI hiring gender gap.
LinkedIn’s 2026 labor market research found women made up just 26% of hires into roles requiring AI skills in the United States. In non-AI roles over the same period, women were about 50% of hires. The pay difference is not marginal. Fortune, reporting the findings at the end of August, put AI roles at around $177,000 against roughly $80,000 for non-AI roles, a gap approaching $100,000 a year.
The word “shortage” is doing a lot of work
The usual explanation is supply. Not enough women studying the right subjects, not enough in the pipeline, not enough applying. Writing in Forbes, Caroline Fairchild pushed back hard on that framing and called the result a design flaw rather than a shortage. The distinction matters more than it sounds.
A shortage is somebody else’s problem to fix, eventually, upstream, in schools. A design flaw is a decision being made right now by whoever writes the job description and runs the interview loop. LinkedIn’s own pipeline analysis found women’s representation drops at several separate points, not one. Women hold about 20% of head of AI roles, 26% of director of AI roles and 18% of technical staff positions. Across 27 countries, women hold roughly 13% of C-suite leadership at AI companies. A single leaky point would suggest a pipeline problem. Leaks at every stage suggest the structure.
Where the AI hiring gender gap is built in
Part of what makes this structural is how quickly the category was invented. AI job postings on LinkedIn have roughly doubled since 2023. Roles built that fast tend to inherit their requirements from whoever already holds similar jobs, which in this field skews heavily male. The requirements then describe that incumbent rather than the work.
Coverage in Axios and CBS News landed on the same conclusion from different angles: the AI jobs boom is redistributing income upward, and women are not in the rooms where that redistribution is happening. Allwork described a triple penalty, where access, perception and progression each take a cut. None of those three are about capability.
Why this reaches beyond big tech
It would be easy to file this under large employer news. It is not only that. The tools being built inside those teams end up running payroll, screening resumes and scheduling shifts at businesses far smaller than the ones writing the code. Fairchild notes a lawsuit alleging AI-assisted layoffs disproportionately affected women who were on leave. Systems inherit the assumptions of the people who build them, and a 26% hiring rate is an assumption about who builds them.
There is a real opportunity here too, and it deserves saying plainly. AI skills are the fastest route to a significant pay increase available to most workers right now, and the required skills are learnable without a computer science degree. That is genuinely open in a way many previous booms were not. The constraint is not talent. It is time, access to the work, and whether anyone in a hiring position treats an adjacent background as qualifying rather than disqualifying. Women often need space to learn more than they need another encouraging speech.
One thing that moves the number
For anyone hiring, including a small business owner filling one role: read the requirements list and cut every item that describes a person rather than the job. Years with a specific framework, a degree subject, a named tool. Most of it is inherited furniture. What remains is usually what the role actually needs, and it widens the applicant pool immediately.
For anyone applying, the more useful question is not whether the requirements are all met. It rarely is for anyone. It is whether the work itself is doable, which is a different question and usually has a better answer. The same hesitation that stops many women using AI openly at work shows up again at the point of applying, and it is worth naming for what it is.
Twenty-six percent is not a discovery about women. It is a measurement of how a new category of work was built, while it was being built. Which means it is still being built.
