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 research reports landed inside three days this month, and read together they name something a lot of women running businesses have probably felt without having words for it. The tool itself is no longer the barrier. Access is close to equal now. What stays unequal is everything built around the tool: who gets taught to use it, who gets asked what it should do, and whose opinion counts as expertise. That is the women AI training gap, and it is now wider than the usage gap it replaced.
The usage gap almost closed, and that deserves saying first
Pew Research Center published a closer look at its gender data on October 16, drawn from interviews with 5,119 US adults. Half of men (50%) and nearly as many women (47%) have used an AI chatbot. ChatGPT use came in at 44% for both groups, identical. Two years ago Pew’s split was 39% against 28%, though it cautions that wording and sample changed between surveys, so that is a direction rather than a clean trend line. Two years is fast, and the progress is real. Women did not sit this one out.
Where the women AI training gap actually shows up
The Shorenstein Center at Harvard Kennedy School published “Are Women Being Left Out of AI?” on October 14, co-led by Nancy Gibbs and Laura Manley. Of the women surveyed, 64% reported no AI training at work, against 49% of men. Nearly half, 46%, said they never use AI at work at all, compared with 36% of men. Women also reported less encouragement and less mentorship around using it. ISACA’s 2026 Tech Workplace and Culture survey, reported October 16, found the same shape in who decides: 53% of women felt they had sufficient authority over company strategy, against 66% of men.
Now set Pew’s confidence numbers beside those. Only 15% of women called themselves extremely or very confident using chatbots, against 22% of men, and 25% said chatbots helped their productivity, against 35% of men. Those get read as a confidence problem. They look more like a downstream number. Less training, less encouragement, less authority over the decision, then a lower self-rating. The order matters, because one of those an employer can fix and the other no pep talk will.
The gap almost nobody measures is in who counts as an expert
The Shorenstein researchers also analyzed over 70,000 news articles about AI from the past year. Men were named in AI headlines 14.6 times as often as women, and cited as experts 4.3 times more often, even after executives, founders, officials and investors are stripped out. Manley’s summary is that women are “rarely quoted as experts.” Respondents of both genders agreed men’s perspectives are better represented in how AI is developed.
That is a control constraint, not a skills constraint. It shapes what gets built, and whose work the next release speeds up. Role models are the easy part. Being in the room where the tool gets specified is the hard part.
Skepticism with reasons attached
Pew found 68% of women saying AI is developing too quickly, against 58% of men. Morgan Stanley’s late September poll of 2,000 US consumers, reported by Axios on October 8, put net AI sentiment at plus 30% among men and minus 1% among women.
Read as mood, that is reluctance. Read against the labor data, it is arithmetic. The National Partnership for Women and Families calculated in April that women are about 47% of all workers but 83% of those in the 15 occupations with the highest AI exposure and the least adaptive capacity, roughly 6 million women in mostly clerical and administrative work. The same brief cites research finding the competency penalty for women who use AI runs at more than double the penalty for men. Women are not misreading the risk. They are judged harder for using the thing they are told they must adopt. Caution that turns out to be well founded is still expensive, though. Being right about the risk does not get the invoicing done.
For an owner with no employer to train her
Those statistics assume somebody else is responsible for the training. For a woman running her own business, nobody is. Professor Christine Woods of the University of Auckland, writing on October 19, argues AI genuinely lowers entry costs, drafting the branding, market research and financial modeling that used to need paid help or the right contacts. She names two limits. Saved time gets absorbed by more email and higher expectations, and the outputs reflect training data dominated by male, Western, well funded entrepreneurship. It will answer confidently either way.
The Cherie Blair Foundation for Women, with Intuit and the World Bank, surveyed over 3,000 women owners in low and middle income countries and found use jumping from 38% to 82% in a year. Depth did not follow. Among frequent users, only 33% apply AI in operations and 35% in bookkeeping and finance. Adopted, not embedded, is the report’s own phrase.
One small thing, if the hours are not there for a course: take one recurring task this week and find the explanation from someone whose business actually resembles yours, not a general tutorial. If anyone else works in the business, write down who was shown the new tool and who was not. That list is usually shorter than expected, and it is the part an owner controls outright.
The adoption numbers will keep climbing and keep being quoted as progress. The number that has barely moved is 4.3. Until women are routinely asked what these tools should do, the tools keep being built to the specification of the people who get asked, which leaves a question no adoption chart answers: useful for whom?
