This study investigates how contemporary artificial intelligence image generation systems interpret and reproduce Indian cultural elements through a comparative analysis of three major platforms: Stable Diffusion, Flux, and Midjourney. While these systems have demonstrated remarkable technical capabilities, their handling of non-Western cultural elements remains understudied. We present a novel methodological framework that combines visual social semiotics with digital anthropology to analyse AI-generated images across multiple dimensions, including representational accuracy, cultural sensitivity, and power dynamics. Our systematic analysis of images generated through increasingly sophisticated prompts reveals significant patterns in how these systems mediate cultural representation. Results indicate that while platforms exhibit varying technical proficiencies, they consistently demonstrate concerning biases in human representation, particularly in their treatment of gender, class, and ethnic identities. The analysis reveals systematic simplification of complex cultural elements and the persistence of orientalist perspectives, despite advances in technical capabilities. These findings suggest that improved technical sophistication alone is insufficient for authentic cultural reproduction; rather, fundamental reconsideration of how these systems process and understand cultural information is necessary. This research contributes both theoretical insights into digital cultural representation and practical implications for developing more culturally sensitive AI systems, while highlighting crucial areas for improvement in the technical architecture of image generation models.