AI-generated CVs reflect deep-rooted gendered ageism, study finds

Artificial intelligence tools like ChatGPT are reflecting and potentially reinforcing biases around age and gender in the workplace, according to a new study published on Nature on October 8.

Researchers found that when asked to generate resumes, ChatGPT consistently made female applicants an average of 1.6 years younger than male applicants. These younger, often less experienced female resumes were then ranked lower by the bot itself, exposing a feedback loop of bias, The Science News reports.

“It’s a notable and harmful thing for women to see themselves portrayed … as if their lifespan has a story arc that drops off in their 30s or 40s,” says Danaé Metaxa, a computer scientist at the University of Pennsylvania who was not involved in the research.

Conducted by computational social scientist Douglas Guilbeault and colleagues from Stanford University, the study used a combination of image analysis, text analysis, and a controlled experiment. The team analysed 1.4 million images and videos, revealing a consistent trend: women were perceived as younger than men across professions, especially in high-status roles like CEOs and doctors.

"These findings could explain the persistence of the glass ceiling for women," says study coauthor and computational social scientist Guilbeault. 

In text data, roles like "secretary" or "intern" skewed young and female, while "chairman" or "director of research" skewed older and male.

In a follow-up experiment, over 450 participants were asked to upload images of workers in various professions and estimate their ages. Uploading pictures did influence beliefs, the team found. Participants who uploaded pictures of female employees, such as mathematicians, graphic designers or art teachers, estimated the average age of others in the same occupation as two years younger than participants in the control condition. Conversely, participants who uploaded the picture of male employees in a given occupation estimated the age of others in the same occupation as more than half a year older.

The team then prompted ChatGPT to generate over 34,000 resumes—one male and one female name for each of 54 occupations. The resumes for women were not only younger but also scored lower in quality, reinforcing bias in hiring pipelines.

Guilbeault notes, “Organizations that are trying to be diverse … hire young women and they don’t promote them.”

The issue isn’t isolated. Even resumes of young men were rated lower than those of young women, suggesting the model penalises youth in men and age in women.

In an accompanying perspective article, sociologist Ana Macanovic of European University Institute in Fiesole, Italy, warns that these AI biases are likely to intensify with widespread use of AI.

Guilbeault adds that efforts to fix bias often focus on single categories like race or gender, but real-world discrimination arises from “the combination of inequalities.” Without a broader approach, AI risks amplifying harmful stereotypes—making the glass ceiling even harder to crack.

By Sabina Mammadli