Self-Reported Generative AI Literacy among Undergraduates in Shandong, China: Dimensional Patterns and Educational Implications

University responses to generative artificial intelligence (GenAI) require evidence about the specific capabilities students perceive themselves to possess. This cross-sectional study examined overall, domain, and item-level self-reports among 616 undergraduates recruited through 18 universities in Shandong, China. A contextually adapted 12-item measure assessed awareness, use, evaluation, and ethics. Descriptive statistics and response distributions were interpreted alongside internal-consistency and confirmatory factor analyses. Overall literacy averaged 3.495 on a five-point scale (SD = 0.650); domain means were 3.688 for ethics, 3.477 for awareness, 3.423 for evaluation, and 3.392 for use. Cronbach’s alpha was .900 overall and .821–.874 across domains. Positive-direction responses accounted for 61.2%–64.9% on ethics items, 45.9%–48.1% on use items, and 45.5%–50.8% on evaluation items. Midpoint responses represented 37.5% for proficient academic use and 38.8% for evaluating capabilities and limitations. Item distributions thus qualified the positive overall mean by locating specific areas of more reserved self-appraisal. The findings support connecting practical application with evaluation and examining how ethical commitments translate into academic decisions. They provide contextual evidence for planning follow-up assessment, while voluntary recruitment and self-report measurement limit population and proficiency claims.