AI, stock photography, and image banks: gender biases and stereotypes

Abstract

Purpose. This study explores the prevalence of gender biases, stereotypes, and representational disparities in stock image banks, contrasting traditional photography platforms with AI-generated visual content. The research aims to assess whether AI mitigates or perpetuates existing biases and stereotypes in visual representation.

Methodology. A case study approach was adopted, analyzing 600 images from four platforms: Shutterstock, Getty Images (traditional stock), Lexica (Stable Diffusion), and Adobe Stock (AI-generated). Standardized prompts, such as “Photography of a smiling person in [location],” were used to ensure comparability. A systematic framework evaluated parameters like gender, age, ethnicity, and the presence of stereotypical elements, revealing trends across platforms.

Findings. The findings confirm persistent biases in both traditional and AI-generated platforms. Traditional stock banks overrepresent women, while AI platforms achieve closer gender balance. Ethnic representation remains heavily skewed toward Eurocentric and Caucasian archetypes, with AI showing slight improvements in Afro-American representation. Age portrayals vary, with AI favoring younger demographics and traditional platforms emphasizing adults. Notably, no images depict individuals with disabilities, highli-ghting a significant gap in diversity. Stereotypes related to beauty standards, such as the use of makeup and accessories, and leisure activities dominate, with minimal representation of professional or diverse cultural roles.

Value. This study provides a comprehensive comparative analysis of traditional and AI-driven stock imagery, highlighting both the limitations and potential of AI to address biases. It contributes a systematic framework for evaluating diversity and representation, offering critical insights for fostering inclusivity in visual media.