AI-Powered Tutoring Quality and Perceived Mathematics Learning Outcomes among Higher Secondary School Students

AI Tutoring has the potential to facilitate learning in mathematics, including through the quality of instruction, perceived usefulness, engagement, and confidence. This paper analyses these associations on the basis of provided partial least squares structural equation modelling result of 470 records. Perceived AI tutoring quality, perceived usefulness, mathematics learning engagement, mathematics self-efficacy and perceived mathematics learning outcomes are reflected in five reflective constructs and twenty-five indicators. The analysis records 5,000 bootstrap samples and tenfold repeated prediction. The suggested seven direct relationships are all positive with confidence intervals not equal to zero. The quality of tutoring is linked with usefulness, engagement and self-efficacy. The engagement is linked to self-efficacy and perceived outcomes and self-efficacy is linked to perceived outcomes independent of engagement. The model is used to explain 39.5 percent of the perceived outcomes. The indicators loading, internal consistency, convergent, and discriminant validity are reported to fulfill the assessment criteria as stated in the workbook. However, prediction is above the linear standard of nineteen of twenty endogenous indicators, but the differences are minor. These results set engagement and self-efficacy as complementary explanatory paths. Nevertheless, an empirical sample cannot be verified due to hypothetical platform labels and missing fieldwork documentation. The results thus detail the provided records of analysis and are not able to determine causal achievement gains. Substantive educational conclusions will still require independent evaluation and validated recruitment documentation.