The paper explores the possible operationalization of the concept of “Otherness”, by testing how effectively Large Language Models can identify its representation in literary texts. Using Joseph Conrad’s Heart of Darkness as an initial case study, we compared the performance of GPT-5 and LLaMA 3 against a human-curated benchmark. Initial results suggest that prompt engineering is key to helping models distinguish between neutral passages and those representing the "Other." This study constitutes the groundwork for a wider project, which aims to scale this methodology across a vast corpus.