Gender gap in AI did not close, it changed shape

Elif Güvençer · 31 July 2026

I was curious about the gender divide in AI. So I dug a little.

Pew's February survey shows 50% of men and 47% of women say they've used AI. Two years ago the same survey showed 39% and 28%. At topline level the gap has closed. Which looks like good news.

When you look at the tools, the picture changes.

Goodwater's 2026 consumer survey puts Grok's user base at 77% male and 63% high-income. Claude's at 57% male, and the most likely of any platform to be paying. ChatGPT, in their words, has the most balanced demographic profile of any platform, at 45% male.

Forbes, referencing the Goodwater data, shows 56% of male AI users pay for subscriptions vs 42% of women. So I would not conclude the gap has closed, I would say the gap has changed. It is not pure access but the tier of access.

The intent with which AI is being used is also different between men and women. 44% of men use AI for work and productivity against 37% of women. Among users, women use it more for healthcare and wellness, cooking and meal prep vs men.

Hilke Schellmann, associate professor at NYU's Arthur L. Carter Journalism Institute, calls this the difference between work that accrues as career capital and work that stays life admin. Both take skill, judgment and a fair amount of trial and error. Only one gets read by an employer as strategic fluency. And when time comes for career advancement, guess which one would count?

The exposure data points the same way. International Labour Organization found 29% of female-dominated occupations exposed to generative AI against 16% of male-dominated ones. The report gives three reasons for that: women are overrepresented in jobs most susceptible to automation; they remain underrepresented in AI-related and science, technology, engineering and mathematics (STEM) occupations; and AI systems themselves often reflect and reproduce the gender biases embedded in societies. These are task-overlap stats not actual job losses but I find the numbers quite telling.

A study out of the Open University, run with Nokia Bell Labs and Politecnico di Torino, shows an interesting divide on the men/women exposure. In male-dominated occupations the exposure sits mostly at the top of the skill and pay distribution, so the more senior you are the more you are exposed. In female-dominated occupations it runs across the whole range so you are exposed throughout the entire career ladder.

High-skill, relatively well-paid roles including nurse practitioners, physician assistants, psychologists, and therapists are among the roles facing the most intense current AI exposure. Administrative and support roles with very high female concentration, including legal secretaries and medical secretaries, face similarly high current exposure. These roles have historically provided accessible entry points into legal and healthcare careers for women without professional qualifications. If AI erodes them without targeted transition support, it narrows those pathways in ways that will compound occupational segregation over time. The report does not go into minority exposure within genders but I imagine that ‘cut of the cut’ of the data would probably show even a more exposed situation...

Responsible and ethical AI by design, policies that can address and help minimize this exposure are of course the correct, systemic way to go. Until they mature, personal agency is going to be needed to make sure everyone but especially women can tap into the potential benefits AI can offer in the workforce. Here is where I would start:

Move off the free default subscription if you can. Pick one tool and go deep, rather than sampling four at surface level.
If you’re using AI for life admin, find the work parallel. Meal planning = scheduling under constraints.
Build something yourself. Start with the thing that annoys you most at work and see whether you can automate it.
You don’t need to know how to build things. Use AI to teach you how to build things.
If you are going to job interviews now, have up your sleeves a few cases you used AI and solved a problem/a mundane task etc.
Basis of value is changing for majority of the disciplines. Spend at least 10% of your time not on a tool but thinking about how your role is changing and where is the new value in the new work system restacked by AI.
Speak up about your AI use publicly and get into conversations with others. It is the fastest way to learn, and the fastest way to be seen as someone who works this way.

On the point about speaking up. Catherine Arrow owns Sensemaking AI, and she's rounding up perspectives from women working in AI and communications. I contributed a piece on The Two Clocks, my framework for how communications leaders should think about AI strategically rather than as a tooling question, and reposition themselves before someone else does it for them. There are several other videos unpacking a different angle on AI. It is free and open to anyone.

The collection is being built regularly. If you work in this space and have a perspective, come forward.

Authenticity statement

I use AI mostly as an interviewer and then an editor. I ask it to interview me on a topic, help me flesh out my thoughts and stress-test ideas, and define the skeleton of what I want to write. It can also suggest angles based on my past writing and what it knows about me. I use it to tear my claims apart and poke holes in my thinking before anyone else does. I have a set of reviewer agents. To name a few: a CCO who reviews whether what I write resonates, and an AI expert who makes sure I stay technically accurate. I speak three languages and English is my second, so I use AI to check my writing is correct and reads well. I use it for image and video creation too.

The irony is that AI disclosure is demanded but the second it is provided, it shows to decrease the trust in the work and/or person. This is a new bias we are up against. If reading this statement changed how you felt about the newsletter, that is the bias doing its work. First step in working with it is awareness! If you are curious about the AI disclosure trust penalty, check out this post.