Identical AI Resumes: Women's Labeled 'Weak' More Often Than Men's

Two identical resumes. Same qualifications. Same AI-generated content. The only difference: one belonged to Emily Clarke, the other to James Clarke. Reviewers questioned Emily's trustworthiness 22% more often than James's, and her competence twice as frequently. The documents were distributed to 1,000 British adults who knew the resumes had been created with artificial intelligence assistance.

The study, conducted by Zehra Chatoo, founder of thinktank Code For Good Now, exposes a troubling pattern in how AI-assisted work is evaluated. When men use AI, reviewers question their effort. When women use AI, reviewers question their integrity. That distinction fundamentally alters the perceived risk of using the technology.

Some feedback on Emily's resume read: "She can't even write a CV herself—not sure she has the skills to carry out the job." James's identical resume received a different response: "He just needed a bit of help putting it together." The physical act of clicking through an application portal feels the same for both candidates, yet the human judgment at the other end diverges sharply.

According to the Fortune report, Gen Z men proved particularly harsh critics. Of their responses, 3.5 times more Gen Z men described Emily's resume as "weak" compared to James's. James's resume achieved a 97% approval rating; Emily's, with identical content, was rated strong by only 76% of respondents.

This isn't an isolated finding. Research from the University of Washington found large language models favored white-associated names 85% of the time versus Black-associated names 9% of the time, and male-associated names 52% of the time versus female-associated names 11% of the time. The systems never preferred Black male-associated names over white male-associated names.

Independent analysis from Brookings Institute confirms that 98.4% of Fortune 500 companies now leverage AI in hiring processes. One company reportedly saved over a million dollars in a single year by incorporating AI into interviews. The efficiency gains are real (and the cost savings are tempting, to be honest).

Women's hesitation to adopt AI tools may be rational rather than risk-averse. A January study from Caltech surveyed 3,000 people and found women were consistently more skeptical than men that AI benefits would outweigh its risks. Their concern has data backing it: a Brookings study found that of roles with high AI exposure but low capacity to adapt, 86% were held by women.

Harvard Business School Associate Professor Rembrand Koning identified this concern in a working paper, noting women face greater penalties in being judged as lacking expertise. "They might be worried that someone would think even though they got the answer right, they 'cheated' by using ChatGPT," Koning explained.

The technical reality is that these systems are often proprietary black boxes. Independent researchers lack access to audit them properly. When companies deploy AI at scale without transparency, they replicate historical discrimination patterns embedded in training data. The algorithms don't create bias from nothing—they amplify what already exists in the human-generated data they consume.

Chatoo's conclusion cuts through the optimism: "If people believe they will be judged more harshly for using AI, they are less likely to adopt it—regardless of their capability." Closing the AI adoption gap means addressing not just how people use AI, but how that use is evaluated.

Whether employers will actually change their screening practices remains the real question. The technology will keep improving, the bias will keep showing up, and candidates will keep clicking submit on applications they know might be judged unfairly.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn