"Beautiful, young, white, and blonde". How Image Generative AI Reinforces Harmful Gender Norms

Image generative artificial intelligence (AI) platforms represent a new, and sometimes troubling, lens for the articulation of gender. Text prompts such as an attractive woman with long wavy hair […] glamorous pin-ups, perfect anatomy, perfect posture (Midjourney, 2023) creates images for users’ consumption at a rate of 34 million per day (Attie, 2023), indicating that this technology is poised to play a dominant role in the transmission, maintenance, and intensification of gender stereotypes.

Employing a critical stance, corpus linguistics software AntConc (Anthony, 2024) was used to analyse users’ text prompts, focusing on gendered, gender-neutral, and gender-connotative noun forms alongside adjectival collocates. Subsequently, a social semiotically-informed multimodal analysis was used on selected images, presented as a case study to exemplify and highlight the corpus linguistics findings.

Bathes described photographs in the media as having “passed through the filter of culture” (1981:16). While generative AI has precipitated a transformative shift in our communication landscape, it appears this cultural filter is still being applied. Critical analysis of this issue places sociolinguists and scholars from other interested fields at the forefront of discussions aimed at identifying and mitigating the propagation of gender hegemonies via image generative AI.