In the study of intertextuality, especially when dealing with sacred texts, the identification of references often relies on an intricate interplay between language, context, and tradition. With the advent of machine learning, new tools have emerged to automate this process, ranging from string-based lexical approaches to more sophisticated models like word embeddings and large language models (LLMs). This paper explores the evolution of these technologies, focusing on the promise and pitfalls of AI-driven methods in identifying intertextual references within ancient texts. While these tools offer unprecedented scale and efficiency, they come with a fundamental question: can we trust them? The lack of explainability in many machine learning models means that using them is, in essence, an act of faith. By drawing a parallel between the interpretive act of engaging with sacred texts and the trust we place in computational models, this paper aims to critically assess the role of AI in textual analysis, questioning whether we are, in fact, placing our faith in the "verbum ex machina".