A 2026 study found that when both women and men create identical resumes with AI assistance, women are seen as less competent and reliable, while men are seen as more proactive. This ability penalty helps explain why women have lower rates of AI usage than men.
getty
Why are women 25% less likely to use artificial intelligence tools than men? A new study debunks the idea that the gender gap is primarily due to women’s lack of AI skills, interest, or access. Rather, women’s reluctance to use AI is a rational response to the competency penalties they face when using AI in the workplace.
The latest evidence on the competency disadvantage faced by female workers when using AI comes from a 2026 study by Zehra Chatoo, founder of Code For Good Now and former strategist at Meta. The study found that women who submitted AI-assisted resumes were rated as less competent and less trustworthy than men who submitted identical AI-assisted resumes. For the evaluator, the female AI uses the signaled incompetence, and the male AI uses the signaled initiative.
This study confirms previous research that shows competency disadvantages for women who use AI to perform their jobs more efficiently. This research shows that organizations cannot close the AI gender gap simply by investing in digital skills training and AI incentives for their employees. Employers must also address gender-based perception bias related to the use of AI.
2026 investigation reveals ability punishment for female AI users
In a new study, Chatoo created AI support resumes for marketing roles and asked 1,000 UK adults to rate candidates during April 2026. Evaluators received identical resumes and were told that the candidate had utilized AI assistance. The only difference in the resumes was the candidate’s name. Half of the raters looked at Emily Clarke and the other half looked at James Clarke.
Despite identical resume content, evaluators rated female candidates much more harshly than men regarding their use of AI assistance.
Evaluators who thought an AI-powered resume was written by a woman were twice as likely to question the candidate’s competency. “She can’t even write her own resume. I don’t know if she has the skills to do the job,” said one of Emily’s resume evaluators.
In contrast, evaluators who attributed an AI-assisted resume to a man were twice as likely to see the candidate as demonstrating initiative. In other words, women’s use of AI was seen as evidence of incompetence, while men’s use of AI to create identical products was seen as a pragmatic problem-solving.
“The identified judgment gap is not just that women are judged more harshly in professional evaluations; that finding is already well-documented,” Chathu said. “That means women who use AI will be judged more harshly specifically for using it.”
Raters were also 22% more likely to doubt a candidate’s credibility if the AI-powered resume came from Emily rather than James. “When men use AI, we question their efforts. When women use AI, we question their sincerity,” Chathu said. “That difference changes the perception of risk of using AI.”
The study also found that even if raters themselves are familiar with AI tools, gender bias does not rule out when evaluating others’ use of AI. Older raters were shown to have less gender bias than male Gen Z raters, who are more likely to use AI themselves. Among Gen Z men, 97% rated James a “strong” candidate, compared to just 76% who rated Emily as a “strong” candidate, a 21 percentage point difference between men and women.
Chatoo warns that the penalties for using AI could be even greater for older women, women of color, or other women who face different sources of bias. “This study focused on gender as a single variable,” Chathu said. “Race, ethnicity, age, and socio-economic background all interact with gender to produce even more differentiated outcomes, and the penalties for AI decisions may be even more severe across these dimensions.”
Women are aware of the AI competency penalty.
Chatoo’s research supports previous research documenting the performance disadvantages faced by women who use AI in the workplace. Women are aware of this gender bias, which reasonably contributes to lower rates of AI usage.
“Women’s hesitancy is not due to a difference in skills; it is an accurate reading of an unequal environment,” says Chathu. “The logical response is to be cautious when the same output is evaluated differently based on its leading name.”
A 2025 study found that women face performance penalties when using AI in the workplace, even in companies that actively encourage employees to use AI tools.
The pilot study was launched after a global technology company discovered that only 31% of its female software engineers were using AI tools, despite the company’s year-long campaign to encourage AI adoption. Company leaders contacted researchers at Peking University and Hong Kong Polytechnic University to investigate the disappointing results.
To assess why female engineers use AI at a lower rate than their male counterparts, researchers asked 1,026 software engineers at the company to evaluate the same computer code. All of the engineers reviewed the exact same code, but received different instructions about whether the code was written with the help of AI and whether the programmer was female or male.
As expected, reviewers rated the objective quality of the same computer code similarly across conditions. However, reviewers gave very different evaluations of the engineers who wrote the code. The reviewers imposed a competency penalty on all engineers who allegedly used AI compared to non-AI users. And the penalties for using AI were much harsher for women than for men.
Male AI users had 6% lower competency ratings than non-AI users, while female AI users had 13% lower competency ratings, even though they all wrote the same computer code. Male raters who did not use AI received the most severe gender-biased performance penalty. Male non-AI users rated female engineers who used AI 26% more harshly than men who used AI to create the same deliverables.
This study further reveals how women’s use of AI reduces their contribution in the workplace. When asked to estimate the relative contributions of engineers and AI tools, evaluators assumed that the AI tools did more work if they considered the programmer to be a woman rather than a man. “AI assistance is framed as ‘evidence’ of AI inadequacy rather than evidence of strategic tool use,” the researchers concluded.
The company’s female engineers were aware of this gender-based competency penalty, which was one of the reasons they were reasonably hesitant to use AI. A follow-up survey of 919 engineers at the company found that women were more likely than men to express fear that their bosses would think less of them as a result of using AI, a valid concern.
How organizations can reduce gender bias in AI use
“The fear of women being judged for their use of AI is not a perception,” Chathu said. “It’s measurable. It’s real.” This means companies won’t close the gender gap in workforce AI adoption with technical training and incentives alone. Employers also need to convince women that they will be evaluated fairly by using AI.
“You can’t upskill people from structural biases,” Chathu said. “Closing the AI adoption gap means addressing not only how people use AI, but also how that use is evaluated.”
Three practices can help organizations reduce the performance disadvantage women face from using AI in the workplace.
1. Collect demographic data about employee AI use and ask why.
“If your AI adoption data isn’t disaggregated by gender, you can’t see the gaps, let alone address them,” Chatoo said. “Start measuring.” AI usage metrics can also be disaggregated by race, age, and other statuses that introduce competency bias.
While collecting AI adoption data is an important step, employers also need to avoid guessing about the causes of usage gaps. Anonymized surveys can help identify the extent to which competency penalty concerns are contributing to lower AI usage among specific workers.
2. Use AI to evaluate deliverables instead of humans
An effective way to reduce gender-based competency bias when evaluating AI-assisted work is to use a “blind review” process. This approach prevents the evaluator from knowing which employees manufactured the product being evaluated. Removing personally identifying information about gender and other potentially biasing characteristics can improve fair and consistent performance evaluations of AI usage.
Even if a blind review process is not feasible, bias may be reduced by instructing raters to rate work products rather than rate workers. In a 2025 study of AI-assisted computer codes, raters rated the quality of the same code equally accurately, regardless of the gender of the coder. Biases against female AI users emerged only when raters were asked to rate programmers’ abilities and contributions.
This finding suggests the importance of training managers to assess the quality of AI-driven work products, rather than assessing worker qualities using AI tools.
3. Use objective metrics
Gender bias is more likely to occur when raters use subjective evaluation criteria such as ability, strength, and reliability. Employers should base hiring evaluations on the specific skills needed for the actual job, and performance evaluations should be based on objective measures of productivity and the quality of the employee’s work product.
These three practices should help organizations not only reduce performance penalties for women using AI, but also promote fairness in any context where stereotypes can bias workplace evaluations. Ensuring fair evaluation of employee AI use can also encourage employee adoption of AI tools.
