From search to imaginary: the algorithmic construction of the refugee figure and the reproduction of bias in the era of generative AI. The Spanish case study of Google AI overviews and Gemini

Interest in refugees and emigration in general has increased over the last decade, becoming a central topic of political and social debate in many Western countries. However, this process cannot be understood without the creation of a socially shared narrative and imaginary around refugees and emigration. In this regard, although many studies have analysed this process, the role of AI in the construction of this imaginary in recent years has not been investigated in depth. The aim of this research is to fill this research gap by examining how Google’s generative AI systems —specifically Google AI Overviews and Gemini image generation—construct representations of refugees through algorithmic content selection and visual synthesis. To achieve this, the study employs a primarily qualitative design, supported by simple descriptive counts of source selection patterns, to analyse textual and visual representations generated by Google’s generative AI systems. Following these analyses, it has been observed that AI-generated images resort to pre-established iconographies in the imaginary about refugees. Furthermore, the analysis reveals how AI Overviews implicitly construct boundaries between “us” and “them” and reproduce previous prejudices and stereotypes with a tendency to simplify diversity.