We cannot deny, for better or for worse, that the democratization and accessibility of artificial intelligence represent an epistemological, social and technical turning point in society. In this context, we are investigating the way in which tools for creating generative images represent and narrate the identities of people from a migrant background. Thus, based on a common list of commands (prompts), entered in distinct geographical areas and in different languages, our aim will be to verify whether the different tools generate representations that corroborate the maintenance of certain paradigms, such as those associated with stereotypes, or whether, on the contrary, they propose another way of seeing the world by breaking with certain "received ideas" about immigrants. To achieve this, we will use three different image-generation tools: Stable Diffusion, DALL-E 2 and Midjourney. This choice is due to the representativeness that this set of tools currently possesses in the market for image generation models. This research will have two main strands: the first, the establishment of a methodological research framework in relation to artificial intelligence tools; the second, the semiotic analysis of the results. The analysis of the images generated will take into account symbolic, political and enunciative aspects, drawing on theoretical approaches from identity theories, sociosemiotics and discursive semiotics.