A woman's viral CV experiment has put Google Gemini at the centre of a fresh AI bias row after she claimed the tool made the male version of her résumé sound more senior, more strategic and more powerful than the female one.
Jennifer Horsburgh said she asked Gemini, Google's AI chatbot, to help update her résumé using her old CV, current career details, case studies and the kinds of roles she wanted. But when she ran the same information under the name 'Jeff' instead of 'Jennifer', she claimed the result was dramatically different.
'I just wanted to update my resume,' Horsburgh wrote on LinkedIn. 'Instead, I accidentally proved how a multi-billion-dollar AI tool hallucinates a glass ceiling for women.'
The post quickly hit a nerve because it tapped into one of the biggest fears around AI and work: that the technology people are using to get ahead may already be carrying the same biases that have held them back.
'Jennifer' Became 'Jeff' And The Language Allegedly Changed
According to Horsburgh, the information she fed into Gemini stayed the same. The only major change was the first name. Yet she claimed the output shifted from softer, more supportive language for Jennifer to stronger, more authoritative wording for Jeff.
In her version of the test, Jennifer's board work was allegedly treated like 'community service', while Jeff was framed as an architect and expert. That difference matters because CVs are not just lists of jobs. They are power documents. The verbs do the selling.
A phrase like 'supported' can make a candidate sound adjacent to impact. A phrase like 'led', 'built' or 'architected' puts them in the room where decisions happen. For women already navigating workplace bias, that kind of wording can quietly shape whether they look like leadership material or someone helping leadership happen.
Horsburgh put it bluntly: 'Same career. Same projects. Same facts. Jeff got the architect. Jennifer got the helper.'
Google has not publicly responded to Horsburgh's specific post at the time of writing. The company's AI principles state that Google aims to use testing, safeguards and monitoring to avoid unfair bias, including bias linked to gender.
Why This Hit Such A Nerve With Women Jobseekers
The reason the post travelled so quickly is that it did not feel abstract. It felt painfully familiar.
Women have long been told to make their CVs more confident, quantify their wins and stop underselling themselves. Now the anxiety is that even when they do all that, an AI tool might still translate their achievements through a softer lens.
That is particularly relevant as more jobseekers use AI to rewrite LinkedIn bios, cover letters, CVs and promotion documents. For many people, Gemini, ChatGPT and similar tools have become unpaid career coaches. But if the advice changes depending on a gendered name, the career coach starts looking less neutral.
This is also where the fashion and creative industries should be paying attention. These sectors run on personal branding, language and perception. A woman described as 'collaborative' may sound nice. A man described as 'visionary' may sound promotable. Same work, different aura, very different outcome.
The issue is not only whether one viral experiment can prove systemic bias on its own. It cannot. The bigger point is that it mirrors a wider pattern researchers have been warning about: AI systems can reproduce, distort or amplify social assumptions already present in the data they learn from.
AI Bias Is Already A Hiring Problem
This is not happening in a vacuum. Research has repeatedly raised concerns about AI and hiring, especially when tools are asked to evaluate, rank or rewrite candidate profiles.
Brookings reported in 2025 that AI résumé screening showed evidence of discrimination based on gender, race and intersectional identity. In its analysis, men's and women's names were selected at equal rates in only 37% of tests. When outcomes were unequal, résumés with men's names were favoured 51.9% of the time, while women's names were favoured 11.1% of the time.
Other research has found that large language models can behave inconsistently in hiring contexts, with preferences sometimes shifting based on names, prompt order or résumé format. That inconsistency is part of the danger. AI can sound calm, polished and objective even when its output is being shaped by messy assumptions.
For women using AI to apply for jobs, the takeaway is not to abandon the tools completely. It is to audit them. Run different versions. Ask for stronger leadership verbs. Strip out gendered language. Compare outputs. Do not assume the first draft is the fairest one.
The viral Gemini test lands because it captures a very modern career fear: what if the glass ceiling no longer starts in the office, but in the prompt box?
