The purpose of this study is to explore the relationship between language, image, and the myths and ideologies that are established through image production using the AI tool Midjourney. This is examined by using as few words as possible as prompts to highlight how Midjourney interprets in the unspoken, in this case, regarding professions and adjectives. Based on Semiotic Theory, Critical media literacy, Ideology critique, and Hegemony, the study provides insights into the use and consequences of this new technology. The method for empirical data collection resembles Grounded Theory, a qualitative research method that involves moving between data collection, analysis, and theory to generate new questions, insights, or understandings that contribute to further data collection. A total of 2924 images have been generated for this study using 731 prompts. The images have been deconstructed and compared, studied from both a micro and macro perspective, and both quantitative and qualitative trends and patterns have been identified. The results of the study show that Midjourney is biased and creates stereotypical representations that carry ideologies and maintain power structures. Generally, it can be said that Midjourney, for all used prompts, tends to align with Western values and perceptions of gender, age, and appearance when these are not specified. This study and its empirical data offer a detailed mapping of how Midjourney creates ideological images. This understanding is necessary to be able to critically work with AI-generated images in school, both as material for analysis and as a tool for creation.