<p>The rapid integration of sophisticated generative artificial intelligence (AI) into mainstream digital life presents escalating challenges to traditional notions of authorship and academic integrity within higher education. While automated AI detection tools have been developed, their reliability is contested, and they often fail to account for the context-specific aspects of student writing. This study seeks to move beyond a purely technological perspective by exploring how university professors identify the distinguishing features of AI-generated texts within student submissions. Employing a qualitative thematic analysis as articulated by Braun and Clarke (2006), this research is based on in-depth, semi-structured interviews with 30 professors from the Faculty of Humanities and Educational Sciences at An-Najah National University in Palestine. The key findings reveal a detection process that synthesizes textual and contextual evidence by combining objective textual indicators with subjective pedagogical assessments. Professors report a reliance on identifying objective markers such as unnatural “linguistic smoothness,” patterns of sophisticated fabrication and citation “hallucinations,” and stylistic redundancy, while also observing the subjective absence of a critical authorial voice. Furthermore, the findings highlight the significant role of subjective contextual intuition and the ethical and emotional challenges educators face. These results carry significant implications for the development of human-informed detection strategies, the urgent need to evolve pedagogical practices to foster critical AI literacy, and the formulation of clear, and fair institutional policies on AI utilization in academic work.</p>