The Operating System · Opinion
Are Women Actually Less Confident About AI, or Have Men Simply Made Themselves the Default Experts Again?
Is the AI confidence gap a women’s problem, or have men made themselves the default experts again?
I have a small theory about men and artificial intelligence.
Apparently, all it takes to become the office AI expert is to be a man who has opened ChatGPT.
I say this as someone who has lost count of the number of times a man has explained AI to me, entirely unaware that I spend my spare time vibecoding for fun. I have listened politely. I have nodded. On occasion, I have wondered whether I should start charging for the privilege of attending these unsolicited masterclasses.
There is something quite remarkable about the confidence with which some men will explain a subject to a woman before establishing whether she knows anything about it. AI has simply given this particular personality trait a new area of expertise.
It got me thinking. Are women actually less confident about using AI, or have men simply made themselves the default experts again?
Confidence is a funny thing
There is research suggesting that the confidence gap is real. A 2023 meta-analysis examined 115 studies across 22 countries and found that men rated their technological abilities higher than women did. The difference was small, but it was there.
In Denmark, a 2025 study of 18,000 workers found that women were 16 percentage points less likely than men in the same job to have used ChatGPT for work. A study of US adults, published in Economics Letters in 2024, found the same pattern: 50% of men had used generative AI, compared with 37% of women. So the researchers asked why. The biggest reason wasn’t age, income or education. It was how much people thought they knew about AI.
Not what they knew. What they thought they knew.
I find that particularly interesting in an industry where confidence has a habit of being mistaken for competence. We all know the colleague who can turn a fairly ordinary observation into a 20-minute presentation, preferably with a diagram. And we all know that being the first person to speak in a meeting does not necessarily mean having the best idea.
Yet somehow, the confidence itself becomes part of the evidence.
A little caution never hurt anyone
Women are often encouraged to be more confident in technology. Apply for the job even if you don’t meet every requirement. Speak up in the meeting. Put yourself forward. Learn the tools. Don’t let imposter syndrome hold you back.
Fine, but notice that confidence is always the thing women are asked to fix. Never the room.
A 2026 study published in PNAS Nexus surveyed around 3,000 people in Canada and the United States. It found that women perceived AI as riskier than men did. The researchers linked this to differences in general risk orientation and women’s greater exposure to AI-related risks. Women were also more likely to question whether AI would deliver positive economic outcomes.
You could call that scepticism. You could call it caution. You could even call it having a functioning sense of self-preservation.
The point is that women are being encouraged to adopt a technology whose risks are still being worked out, while their concerns are sometimes treated as a confidence problem.
I would quite like to see the same enthusiasm for asking difficult questions that we see for announcing AI strategies.
The boys have discovered a new toy
If you want to know why women might hesitate, look at what the tool has already been used for. Some of its most immediate applications have been deeply unpleasant.
Generative AI has made it easier to create fake intimate images of women and distribute them without consent. A 2024 study involving more than 16,000 respondents across ten countries examined non-consensual synthetic intimate imagery. It found that 2.2% of respondents had experienced victimisation, while 1.8% reported perpetration behaviours.
The same researchers also asked more than 16,000 adults in ten countries about sextortion: threatening to share someone’s intimate images. Men were 1.43 times more likely than women to report doing it. And that’s only the men willing to admit it. Surveys like this rely on people reporting their own behaviour, so I suspect the real number is a good deal higher.
There is something grimly predictable about a technology being celebrated for democratising creativity while also making it easier to manufacture sexual abuse.
And before someone tells me that technology itself is neutral, I’m aware. A kitchen knife can also be used to make a sandwich. That doesn’t mean we should ignore what people are doing with it.
Women have good reason to be interested in who gets access to these tools, how they are used and who is left dealing with the fallout.
Who gets to define progress?
Then there is the environmental question.
The International Energy Agency reported that electricity consumption by data centres increased by 17% in 2025. Electricity consumption from AI-focused data centres grew by 50%. Those numbers are worth remembering the next time an app introduces an AI feature that nobody asked for.
The industry is in a race to make AI part of practically everything. Sometimes the use case is genuinely useful. Sometimes it feels as though someone has put a chatbot in a product simply because the board asked where the AI strategy was.
Meanwhile, the infrastructure behind these products requires electricity, water and physical resources. The environmental impact deserves the same scrutiny as the productivity claims.
There is also a wider gender context. An OECD report from 2023 found that women were eight percentage points more likely than men to see climate change as a major threat. This does not tell us whether women are more concerned about AI’s environmental footprint specifically. It does show a difference in how women and men report thinking about environmental risk.
I wonder what AI development would look like if the people asking about its energy consumption, social consequences and potential for abuse had as much influence as the people selling its productivity benefits.
The confidence problem is somewhere else
The advice to women about AI keeps landing in the same place. Be bolder. Stop being so careful. In other words, be more like the man who opened ChatGPT last week and now runs the AI working group.
Women should absolutely have access to AI education, technical roles and the opportunity to build these products. There is no question about that.
But I am less interested in another initiative teaching women how to feel more confident while the industry carries on without them.
I want to know who is deciding what gets built. Who is setting the rules. Who is questioning the assumptions in the models. Who is considering the people whose images, jobs, data and livelihoods may be affected.
And I want to know why a man can spend an afternoon prompting ChatGPT and come away with the confidence of a newly appointed CTO, while a woman who has been working with technology for years is still expected to demonstrate that she belongs in the conversation.
Perhaps women are less confident about AI. Some are more cautious, and some simply have a longer list of questions before they’re prepared to call something progress.
I don’t think the industry should be in such a hurry to correct that.
Sources
- Gender differences in self-assessed technological ability, a meta-analysis of 115 studies in 22 countries: Social Science Research, 2023
- ChatGPT use among 18,000 Danish workers: Humlum & Vestergaard, PNAS, 2025
- Generative AI use among US adults: Aldasoro et al., Economics Letters, 2024
- Women's scepticism toward AI, around 3,000 people in the US and Canada: Magistro et al., PNAS Nexus, 2026 (Northeastern University summary)
- Non-consensual synthetic intimate imagery in ten countries: Umbach et al., CHI 2024
- Sextortion in ten countries: Henry & Umbach, Computers in Human Behavior, 2024
- Data centre electricity use in 2025: International Energy Agency, 2026
- Climate change as a perceived threat, by gender: OECD Skills Outlook 2023