This study looks at cultural and geographic bias in AI-made images, exploring how the text-toimage system Midjourney version 7 shows Ghana and Sweden differently based on learned visual patterns. As AI images become more common worldwide, the ideas built into these systems have important social and political effects. While past research has shown bias based on demographics in AI, the cultural and location-based sides of that bias have not been studied much. This study fills that gap by using Rose's (2016) Visual Discourse Analysis on 32 images created from four similar prompt types: family, busy street, market, and office workplace. The analysis uses the ideaof the Algorithmic Gaze, combining Said's (1978) Orientalism, Graham's (2010) and Kotliar's(2020) ideas about computational power, and postcolonial visual culture work by Hall (1997) and Lutz and Collins (1993).The study concludes that the Algorithmic Gaze does not invent new biases but inherits and automates existing ones, extending the reach of a colonial representational system into new technological contexts at unprecedented scale. The findings contribute to the growing literature on AI bias by demonstrating that cultural and geographic bias is not incidental but structural, and by showing that addressing it requires historical and political engagement with the archives from which AI training data is drawn, not merely technical mitigation.