Abstract
Abstract
Artificial intelligence tools have permeated undergraduate academic life in India, yet their downstream cognitive and academic consequences remain poorly understood among rural students who adopt these tools in the absence of institutional guidance. Grounded in Extended Mind Theory and Cognitive Load Theory, this study develops and tests a sequential mediation model in which AI tool usage (AITU) influences academic performance perception (APP) through cognitive offloading behaviour (COB) and self-regulated learning behaviour (SRLB). Data were collected from 425 rural undergraduate students enrolled in government and private colleges in India. A three-stage analytical sequence, Exploratory Factor Analysis (EFA), Confirmatory Factor Analysis (CFA), and Structural Equation Modelling (SEM) with 2,000 bootstrapped resamples in SPSS 26 and AMOS 24, was employed. Harman's single-factor test confirmed that common method bias was not a dominant concern (single-factor variance = 38.40%, below the 50% threshold). The four-factor measurement model demonstrated acceptable fit (chi-square/df = 2.961, CFI = .925, RMSEA = .068). Structural results confirmed that AITU positively predicted COB (beta = .723, p < .001) and SRLB (beta = .586, p < .001), while SRLB significantly predicted APP (beta = .362, p < .001). The sequential AITU, COB, SRLB, APP indirect pathway was significant (p = .002, BC 95% CI excluding zero), while the COB, APP direct path was not robustly supported in bootstrap estimation. Multi-group SEM revealed that lower-performing students (academic score < 60%) exhibited a substantially stronger AITU, COB relationship (beta = .961) than their higher-performing peers (beta = .681), indicating differential vulnerability to AI-induced cognitive offloading. Findings suggest that unsupervised AI adoption elevates cognitive offloading, which erodes self-regulated learning and ultimately constrains academic performance perception, with academically weaker students disproportionately at risk. Institutions should integrate structured AI literacy curricula and metacognitive scaffolding to convert AI adoption from a convenience habit into a productive learning strategy.