Negotiating generative AI in online and distance higher education: lecturers' task-selective adoption in Yogyakarta, Indonesia

Purpose This study examines how lecturers in Yogyakarta, Indonesia, adopt and negotiate the use of generative artificial intelligence (AI) in academic work, focusing on perceived usefulness, task-selective use, ethical ambivalence and institutional readiness. Design/methodology/approach A convergent mixed-methods design combined an online survey of 120 lecturers from public and private universities in Yogyakarta with semi-structured interviews with eight lecturers. Survey data were analyzed descriptively and compared across age, academic rank and institution type using the Kruskal–Wallis H test; interview data were analyzed thematically. The strands were merged during interpretation. Findings All respondents (100%, n = 120) agreed that generative AI improves academic efficiency (M = 4.47). However, agreement declined across presentations (56.7%), teaching material preparation (53.3%), disciplinary understanding (51.7%), scholarly articles (48.3%), research (46.7%) and academic books (46.7%), indicating task-selective adoption. None of the task-specific usefulness ratings differed significantly by age or academic rank (all p > 0.05), but perceived training need differed significantly by age and rank, with a large effect for age and a moderate effect for academic rank (both p < 0.001). Institution type was associated with attitudes toward AI for academic books and toward AI possibly replacing lecturers (both p < 0.05). Interviews revealed the same task-selective logic, alongside ethical ambivalence about plagiarism and dependency and a perceived absence of institutional guidance. Research limitations/implications The study used purposive convenience sampling in one Indonesian province, with an unequal public-private institutional ratio, a cross-sectional self-report survey and eight interviews. The findings therefore cannot support causal claims or represent all Indonesian lecturers, higher education institutions, or open and distance learning (ODL) systems. Negotiated professional adoption remains an emergent interpretation that requires further testing. Future research should use multisite, more balanced and longitudinal designs to examine institutional differences and changes in lecturers' AI practices as policies and technologies evolve. Practical implications Universities should issue task-specific policies that distinguish legitimate AI assistance from inappropriate substitution, especially in scholarly writing and research. They should establish authorship-disclosure rules, verification procedures and clear academic-integrity mechanisms. Lecturers need guidance on checking AI-generated content and preserving professional judgment. ODL administrators should provide differentiated training based on seniority and need rather than uniform programs. National bodies should set minimum standards, while institutions should adapt operational policies to their own contexts. Social implications Clear institutional guidance can support responsible generative AI use while protecting academic integrity, professional accountability and trust in higher education. Differentiated training can reduce uneven preparedness across age groups and academic ranks. Transparent rules on authorship, verification and acceptable use may help lecturers and students understand the boundaries between assistance and substitution. These measures can encourage fairer, more consistent and more trustworthy use of generative AI in digitally mediated and open and distance higher education. Originality/value The study proposes negotiated professional adoption as an emergent, contextually grounded conceptual interpretation that extends the Technology Acceptance Model with professional, ethical and institutional dimensions specific to generative AI in open and distance higher education.