Women Face Competence Penalty for Using AI at Work, Study Finds

Women are 25% less likely than men to use artificial intelligence tools at work, but the gap has little to do with skills or access. Instead, new research reveals women face systematic bias when they use AI—a competence penalty that makes their reluctance entirely rational.

A 2026 study by Zehra Chatoo, founder of Code For Good Now and former Meta strategist, exposed the double standard by asking 1,000 U.K. adults to evaluate identical AI-assisted resumes during April 2026. The only variable: half saw the candidate named Emily Clarke, half saw James Clark.

Evaluators judged the woman candidate twice as harshly for using AI assistance. Comments about Emily's resume questioned her basic competence: "She can't even write a CV herself—not sure she has the skill to carry out the job." Meanwhile, evaluators viewing James's identical resume were twice as likely to credit him with showing initiative and pragmatic problem-solving.

The bias extended beyond competence. Evaluators were 22% more likely to doubt trustworthiness when the AI-assisted resume came from Emily rather than James. "When men use AI, we question their effort. When women use AI, we question their integrity," Chatoo said. "That difference changes the perceived risk of using AI."

Why it matters

This research reveals that closing the gender gap in AI adoption requires more than training programs or technology access. Organizations investing millions in AI tools may see disappointing adoption rates among women employees not because of capability gaps, but because women accurately perceive an uneven evaluation environment. The findings suggest companies need to address bias in how AI-assisted work is assessed, not just how it's created.

Prior research confirms the pattern

Chatoo's findings align with a 2025 study conducted at a global technology company where only 31% of female software engineers used an AI tool despite a year-long adoption campaign. Researchers from Peking University and Hong Kong Polytechnic University asked 1,026 engineers at the company to evaluate identical computer code under varying conditions.

While reviewers rated the code's objective quality similarly regardless of who wrote it, they imposed harsh competence penalties on the engineers themselves when AI was involved—and the penalty was twice as severe for women. Female engineers using AI received 13% lower competence ratings compared to 6% lower ratings for men, despite producing identical work.

Male evaluators who didn't use AI themselves were particularly biased, rating women engineers who used AI 26% more harshly than men. Evaluators also assumed the AI tool had contributed more when they believed the coder was female, framing AI assistance as proof of inadequacy rather than strategic tool use.

Women engineers at the company were aware of this penalty, which explained their rational reluctance to adopt the tools.

Reducing bias in AI evaluation

Chatoo emphasizes that organizations cannot solve this problem through skills training alone. "You cannot upskill people out of structural bias," she said. "Closing the AI adoption gap means addressing not just how people use AI, but how that use is evaluated."

Three practices can help reduce the competence penalty:

First, gather demographic data on AI adoption rates and survey employees about concerns driving usage gaps. If data isn't disaggregated by gender, the problem remains invisible.

Second, implement blind review processes that remove identifying information about who created AI-assisted work. When blind review isn't feasible, train evaluators to assess work products rather than making judgments about the worker's competence.

Third, use objective evaluation metrics tied to specific job tasks rather than subjective criteria like "competence" or "trustworthiness," where stereotypes flourish.

Chatoo warns the penalty likely compounds for women facing intersectional bias based on race, age, or socioeconomic background. "Women's hesitation is not a skills gap," she said. "It is an accurate read of an uneven environment."

The research was first reported by Michelle Travis in Forbes.