What is said? What is silenced? Representations of Motherhood and Discursive Invisibility in Artificial Intelligence: The Cases of Gaza and East Turkestan

Abstract

This study aims to examine how generative artificial intelligence systems shape gender representations and which experiences they render discursively visible or invisible. In this context, the responses generated by ChatGPT, a U.S.-based large language model, and DeepSeek, a China-based model, regarding motherhood, femininity, and caregiving experiences in the contexts of Gaza and East Turkestan were comparatively analysed. The study was conducted using a qualitative content analysis and comparative discourse analysis approach. Data obtained from a total of 60 prompts were evaluated in terms of mechanisms of discursive visibility, omission, and reframing. The findings reveal that both models produce representations that humanise motherhood; however, these representations vary significantly depending on the context. ChatGPT, particularly in the context of Gaza, tends to make human suffering visible while avoiding explicit identification of the perpetrators of violence and political responsibility, thereby producing a neutralising discourse. In contrast, DeepSeek, in the context of East Turkestan, tends to avoid representation, suppress certain narratives, and at times generate a discourse that may be interpreted as implicitly legitimising.These findings suggest that artificial intelligence systems are not merely tools for information generation but also epistemic actors that selectively construct gender representations. The study contributes to the literature on AI ethics and feminist technology studies by demonstrating that bias in artificial intelligence should be evaluated not only through explicit prejudice but also through processes of discursive invisibility and omission.