Colourism is a positive bias towards people with lighter skin, which manifests itself at the intra- and/or inter-ethnic level (Doudy-Michez, Pohl and Moerenhout, 2024). In Belgium and France, this research is the first to evaluate AI-generated images in relation to colourism among black women according to their hierarchical level in the care sector. Black women are overrepresented in care professions, particularly in subordinate positions (Gatugu, 2017). Among Black women, ethnic group and gender stereotypes, as well as the type of profession, influence the perception of skills (Cuddy, et al., 2008). Generative AI algorithms reproduce the biases affecting people who are discriminated against in the real socio-professional world (Lacroux & Lacroux, 2021).


In the design of this study, 50 prompts were submitted to three AIs: Copilot, Sora (ChatGPT) and Gemini. The prompts included three skin colour types: light mixed race, dark mixed race and black, and the hierarchical level of the position (care assistant, nurse and paediatrician).


Analysis of the images shows that AI fails to take into account colourism affecting dark-skinned mixed-race black women and that there are differences in skin colour depending on hierarchical level. The generated photos show that they have the same skin colour as dark-skinned black women. Our research demonstrates that generative AI must include colourism in order to represent the existing professional realities in the care sector.