This research investigates whether gender and physical appearance influence candidate evaluations generated by a GPT-based artificial intelligence (AI) recruitment system. The study addresses the policy question of whether automated screening tools, often promoted as neutral and objective, may in fact perpetuate visual and gender-based biases, thereby affecting equal opportunity in hiring.
We conducted a large-scale controlled experiment using 600 real job postings, spanning six occupational domains: economics/finance, computer science/data, accounting, education, customer service, and sales. For each posting, six nearly identical CVs were created varying only by gender, and attractiveness (attractive, plain, or no photo). In total, 3,600 CVs were evaluated by OpenAI's GPT-4, prompted to assign interview suitability scores from 1 to 100. Occupational fields were categorized as male-dominated, female-dominated, or gender-neutral; jobs were further classified by required experience, and whether they were public-facing or office-based.
Statistical analysis reveals systematic, context-dependent biases. Attractive women enjoyed a beauty premium in both male- and female-dominated fields, while men benefited only in female-dominated domains and were penalized for including photos in male-dominated fields. Plain female candidates in female-dominated fields faced notable—though not always statistically significant—penalties, suggesting gender-visual stigmatization. Bias effects were strongest in entry-level jobs, where less professional information was available, indicating greater reliance on peripheral visual cues. Contrary to patterns observed among human recruiters, no significant appearance advantage was found in public-facing roles compared to office jobs.
The study extends evidence of labor market discrimination from textual to visual AI bias, showing how stereotypes interact with occupational gender composition. Findings highlight that even advanced AI may reproduce social inequalities with significant implications for hiring practices.