Academic-Integrity Concerns, Satisfaction, and Perceived Learning Benefits of ChatGPT Across 15 National Higher Education Contexts: A Multigroup Study

Generative artificial intelligence is becoming embedded in higher education as students evaluate its usefulness and integrity implications. We analyzed Global ChatGPT Student Survey data to examine associations among academic-integrity concerns, satisfaction, and perceived learning benefits across national academic settings. The source dataset contained 23,218 records from 108 identifiable countries. The primary multigroup analysis included 8650 respondents from 15 countries, meeting study-defined thresholds for FIML-eligible sample size (≥300), complete nine-indicator data (≥200), completeness (≥80%), and model feasibility. A seven-country robustness analysis (N = 5871) applied stricter sample-size and completeness criteria. Configural and full metric invariance were broadly supported; full scalar invariance was not, so latent means were not compared. Satisfaction was positively associated with perceived learning benefits in all 15 countries. Academic-integrity concerns were negatively associated with satisfaction in 14 countries and directly with perceived learning benefits in 13, although five and six respective confidence intervals included zero. Robust Wald tests of unstandardized paths detected heterogeneity, while equality constraints caused only trivial global-fit deterioration. The robustness analysis reproduced the directional pattern and heterogeneity evidence. These findings provide a limited empirical window into students’ GenAI-related academic norms and evaluations; they do not directly measure culture or establish causal mediation.