RDEL #148: How does gender impact adoption of GenAI tools?

Research-Driven Engineering Leadership

SubscribeSign in

RDEL #148: How does gender impact adoption of GenAI tools?<br>Across 100+ countries, men adopt generative AI about 22% more than women, and the gap has stalled rather than closed.<br>Jun 16, 2026

Share

Welcome back to Research-Driven Engineering Leadership. Each week, we pose an interesting topic in engineering leadership and apply the latest research in the field to drive to an answer.<br>Generative AI is meant to be for everyone, but not everyone is adopting it at the same rate. When the gap follows predictable lines, it shapes who captures the productivity gains and who falls behind. This week we ask: how does gender shape who actually adopts and uses generative AI?<br>Thanks for reading Research-Driven Engineering Leadership! Subscribe for free to receive new editions to your inbox weekly.

Subscribe

The context

Generative AI could be a great equalizer. The tools are affordable, broadly accessible, and require no prior skills to try. In theory, they could extend expert-level capability to anyone, including groups that have historically faced barriers to building technical skills. But the history of technology adoption complicates that optimism. New tools tend to diffuse first through the people who already hold technical and managerial roles, then spread outward through peer networks. If certain groups are less exposed early on, they fall behind on familiarity, and that lag can compound.<br>This is not a story confined to specialized software tools. It spans how people use generative AI everywhere, at work and at home, across professions and countries. The same forces shaping who reaches for ChatGPT or a coding agent are at play across the broader workforce, and early fluency is starting to look like a durable career advantage. The question is whether these gaps are real and durable, or just early-adoption noise that evens out on its own.<br>The research

Researchers at Harvard, Stanford, and Berkeley assembled one of the most comprehensive published a working paper that performs a systematic review of 76 academic, industry, and government sources spanning more than 100 countries with SimilarWeb web-traffic data. They reviewed the ten most-visited AI tools, pooling 318,924 respondents from sources that report usage rates for both men and women.<br>Here’s what they learned:<br>The gender gap in AI adoption is large and nearly universal. Pooling across sources, 47.8% of men used generative AI versus 39.3% of women, a relative gap of about 22%. The raw gap favored men in 56 of the 58 sources they plotted.

The gap is shrinking, but it has stalled rather than closed. It has narrowed since ChatGPT launched, but stabilized at roughly 16% since early 2025. In the web-traffic data, women rose from just under 35% of visitors in January 2023 to about 40% by January 2026, then flattened there from late 2024 onward.

Share of AI tool website traffic from women over time

It isn’t explained by men working in more technical jobs. The gap persists inside the same occupation and even the same company. Tracking software engineers at one global tech company after it launched an internal AI coding tool, Gai, Hou, and Tu (2025) found 43% of male engineers but only 31% of female engineers used it at least once, “despite identical job roles and identical access.”

The gap is widest exactly where the frontier is moving. Gaps are smallest for companion-style chatbots and largest for tools like vibe-coding app builders. Among quantitative social scientists, those with typically male names were 144% more likely to use coding agents weekly than those with typically female names.

Female share of US AI tool website traffic over time, by tool

The gap shows up on the intensive margin too. Even among women who adopt, they tend to send fewer prompts, log fewer daily minutes, and have shorter sessions. Across the top ten U.S. tools, the female share of traffic ranged from 26.7% on Grok to 44.1% on ChatGPT, with every tool male-skewed.

The application

The authors group the causes into two buckets: familiarity gaps that the market tends to close as tools diffuse, and stickier institutional and social frictions like confidence, competence penalties, time for upskilling, and trust concerns. Their data suggests both are at work, which means the gap won’t fully self-correct, and the parts that persist are precisely the ones organizational choices can influence.<br>For engineering leaders, these patterns are likely to show up on your team, and can be addressed in the following ways:<br>Invest in structured learning opportunities. Effective AI use takes upfront investment in prompting, libraries, and workflow integration, and that investment is easier to skip when discretionary time is scarce. Fund it directly: protected learning time, hands-on onboarding for new tools, and shared prompt libraries so fluency doesn’t depend on who happens to be...