Generative AI Use and Legal Research Competence among Law Students in Bangladesh: The Roles of AI Literacy and Formal Training

Generative artificial intelligence (GenAI) is rapidly entering legal education, yet evidence remains limited on whether frequent use itself translates into stronger legal research competence in emerging legal-education systems. This study examines the association between GenAI use, AI literacy/responsible-use awareness, formal AI training, and self-reported legal research competence among 385 respondents in Bangladesh. The supplied cross-sectional dataset was analysed using descriptive statistics, item-level diagnostics, bivariate association tests, one-way analysis of variance, and hierarchical moderated regression. The results show that use frequency had a small positive bivariate association with legal research competence (r = .138, p = .0066), whereas AI literacy/responsible-use awareness had a substantially stronger association (r = .353, p < .001). Formal AI training was also associated with higher competence scores: respondents with no training averaged 3.45, those with short orientation 3.56, and those with course/certified training 3.62; the overall difference was statistically significant (F = 5.29, p = .0054, η² = .027). In the multivariable model, AI literacy remained the dominant predictor, formal training retained a positive effect, and the direct contribution of use frequency was no longer statistically significant. The frequency × AI-literacy interaction was also non-significant. These results suggest that competence is linked less to exposure alone than to verification, risk awareness, human oversight, and structured training. However, internal-consistency coefficients for the multi-item blocks were low, requiring cautious interpretation and instrument refinement before confirmatory latent-variable modelling. The study contributes a capability-centred perspective to GenAI adoption in legal education and identifies concrete priorities for curriculum design in Bangladesh.