The present study aimed to investigate the challenges of bilingual students and to compare human and machine analyses of first-grade teachers’ experiences in this context. The research employed a qualitative approach using Colaizzi’s descriptive phenomenological method. Data were collected through semi-structured interviews with 15 first-grade teachers from bilingual regions in Uramanat, Iran. The data were analyzed through two parallel methods: (1) human analysis based on Colaizzi’s seven-step framework, and (2) machine analysis utilizing GPT-4 and the chain-of-thought prompting technique. The human analysis identified four major categories: educational, linguistic, psychological, and social challenges. In contrast, the GPT-4 analysis classified the data into six themes: fundamental linguistic difficulties, negative psychological experiences, cultural alienation, home-school mismatch, educational disengagement, and lack of institutional support. Comparative analysis revealed significant overlaps between the two approaches, though the human analysis, with its deeper sensitivity to cultural context and conceptual integration, provided a more unified interpretation. Conversely, GPT-4 demonstrated high precision in detecting textual patterns and independent themes, thereby unveiling novel dimensions of the problem. The findings suggest that when machine-generated outputs are reviewed by human analysts and evaluated within a theoretical framework, the combination can serve as a valuable complement; otherwise, there is a risk of superficial interpretations.