From Individual Competence to Distributed Capacity: How Islamic Religious Education Teachers Verify Generative-AI Content in Padang, Indonesia

Abstract

Generative artificial intelligence (AI) produces fluent religious exposition without disclosing the ground of its authority. For Islamic Religious Education (in Indonesia, Pendidikan Agama Islam, hereafter PAI), this limitation touches a foundational concern, because the validity of religious knowledge is determined not only by the truth of its content but also by its chain of transmission. Normative scholarship has articulated how tabayyun (verification) and maqāṣid (the higher objectives of the law) should guide the use of AI, and an instrument for Islamic AI literacy has been developed for university students. What has been missing is empirical evidence of how schoolteachers of religion narrate the application of such norms when confronting concrete AI outputs. This study addresses that gap through a reflexive thematic analysis of semi-structured interviews with fifteen senior-secondary PAI teachers in Padang, West Sumatra, sampled purposively for maximum variation in employment status and school sector. Four themes were developed. First, AI shifts the allocation of teachers’ labor from text production to verification rather than reducing the overall workload. Second, the commitment to verify is uniform across the sample, whereas the standard of authority deemed adequate varies sharply, ranging from the classical sanad to personal networks. Third, the limit of delegation is drawn at what cannot be conveyed as text, not at any threshold of informational accuracy. Fourth, the capacity to meet that standard appears to track institutional proximity to scholarly authority. A systematic search of the entire corpus found no Minangkabau adat category, and the repertoire teachers mobilized was normative-translocal; because the interview guide did not prompt for adat, this absence is treated as an observation to be tested rather than as settled evidence. On the strength of the whole sample, the study proposes reframing Islamic AI literacy from an individual competence into a distributed capacity, a hypothesis the data motivate but do not settle, resting most heavily on a single matched pair of cases. The implication is direct: training programs that target individual skills without providing access to scholarly authority risk widening the very gap they seek to close