The rise of artificial intelligence (AI) in textual analysis raises fundamental questions about its ability to grasp the subjectivity inherent in literary works. While AI systems, particularly those based on machine learning, are increasingly capable of identifying linguistic patterns, simulating stylistic analyses, and generating automated summaries, they struggle to grasp the sensitive, emotional, and aesthetic dimensions of literary language. This article explores the limitations of these tools when faced with phenomena such as irony, ambiguity, polyphony, and stylistic singularity. Drawing on digital humanities, literary theory, and computer science, the study highlights the tensions between algorithmic objectification and the subjective experience of reading. In particular, it emphasises the lack of interiority in machines (their inability to feel, remember, or imagine) and the tendency of AI models to reduce literary complexity to fixed analytical categories. In light of these findings, the article advocates for a critical and hybrid approach based on the complementarity between artificial intelligence and human interpretation. Rather than rejecting digital tools outright, it defends their responsible and contextual use, recognising their epistemological limitations in capturing literary subjectivity. Ultimately, this work invites broader reflection on the place of technology in the humanities and on the conditions for a fruitful dialogue between computation and interpretation in literary analysis.