The AI Gender Gap Story Everyone's Telling Is Missing the Point
Harvard Business School published a widely-shared piece this year built on research from Rembrand Koning and colleagues: women adopt generative AI at roughly 25% lower rates than men, across 18 studies and 140,000+ people. The headline explanation the one that's been repeated in every LinkedIn post since, is that women are worried about being judged. That using ChatGPT feels like "cheating," and that admitting to it costs more for a woman's credibility than a man's.
It's a compelling story. It's also, based on what I actually heard when I asked women directly, probably not the real story.
What I did instead of citing the study
Rather than take the headline at face value, I ran a short survey of my own, 20 professional women, mostly 35–54, spanning founders, freelancers, senior leaders, and people mid-career or in transition. Small sample, self-selected, skewed toward women already engaged enough with AI to fill out a survey about it. I'm not claiming this overturns a peer-reviewed meta-analysis of 140,000 people. But it's real data from real women, and it tells a different story than the one getting quoted everywhere.
The gap isn't confidence. It's noise.
Here's what actually came back:
Around half of respondents already use AI regularly and feel confident with it. Most of the rest are using it and actively trying to get better. Only one or two haven't integrated it at all.
Not a single respondent cited fear of being judged, or worry about "cheating," as a reason they hold back.That's the headline explanation in the HBS piece, and it simply didn't show up.
What did show up, repeatedly: concern about accuracy, concern about privacy and security, not knowing which tool to use among an overwhelming number of options, not having time to learn properly, and — notably — worry about becoming too dependent on it.
Read that last cluster again. These aren't the concerns of women avoiding AI out of insecurity. They're the concerns of women already using AI seriously enough to worry about its failure modes. One respondent, a consultant working on AI governance, put it plainly: her hesitation is about responsible use, not personal confidence. Another, a computer scientist, said she has "sound knowledge of its double-edged sword." These aren't women shrinking from the technology. They're women who've looked at it closely enough to be appropriately wary.
What they actually want is not what the "fix" assumes
The HBS piece's prescription is essentially cultural: build psychological safety, normalize experimentation, make people feel it's okay to try and fail. That's not a bad idea in general. But when I asked these women directly what they'd want from a space built to help them navigate AI, the answers were almost entirely practical: hands-on workshops, real-world use cases, help figuring out which tools actually work for their situation, conversations with other women already using AI successfully in their field, and mentoring. Several explicitly said they didn't want anything "too basic", a sign of a population past the confidence-gap stage and into the curation-gap stage.
One respondent even pushed back on the premise of a women-only space entirely, saying she wanted to hear from men too and wasn't fully sold on the framing. That's worth sitting with rather than editing out — it's a reminder that "women avoiding AI" isn't one story, and neither is "what women want to fix it."
The real gap, if there is one
If this small sample says anything, it's this: the barrier isn't primarily psychological, and the fix isn't primarily about permission. It's about time, trustworthy curation in a landscape that changes weekly, and confidence that the accuracy and privacy trade-offs are actually manageable. That's a very different problem to solve than "make women feel less judged for using AI" — and it demands a very different kind of community or resource than a safe-space training program.
None of this means the HBS research is wrong at a population level, a 140,000-person meta-analysis deserves real weight, and a 20-person informal survey doesn't overturn it. But it's a useful reminder that a clean, quotable finding at scale can flatten a much messier, more practical reality on the ground. If you're building anything a product, a program, a community for the specific women you're trying to reach, the population-level story is a starting point, not a blueprint. Go ask them yourself. You might find, like I did, that the thing everyone's citing isn't the thing actually standing in the way.
Based on an informal survey of 20 professional women conducted August 2026. Referenced research: Koning et al., "Global Evidence on Gender Gaps and Generative AI," discussed in HBS Working Knowledge.
