The widespread adoption of generative artificial intelligence (AI) tools in higher education has raised concerns about cognitive offloading and its potential erosion of critical thinking capacities. While prior research has documented negative associations between AI dependence and critical thinking in general student populations, no study has examined this relationship in a chemistry-specific educational context or applied Bloom's taxonomy-based critical thinking measurement. This study investigated the relationships among AI-related cognitive offloading, domain-specific AI use in chemistry education, and multidimensional critical thinking capacity among students and teachers. A cross-sectional survey was conducted with 320 participants (80.3% female; predominantly aged 17–25) during December-January 2026 academic year. Three instruments were administered: the AI Usage and Cognitive Offloading Scale (AICOS; N = 320), the Critical Thinking Questionnaire (CThQ; N = 320), and the AI in Chemistry Education Scale (AICES; N = 305). Confirmatory factor analysis, reliability analysis, Welch's ANOVA with Tukey post-hoc tests, and bivariate correlation analyses were performed in Jamovi 2.5. The CThQ confirmed acceptable structural validity (CFI = .939, RMSEA = .084) in this context, with strong overall reliability (α = .916). The majority of participants (59.1%) demonstrated high-level critical thinking scores. Cognitive offloading was predominantly average (67.8%). No statistically significant correlations were found among AICOS, CThQ, and AICES (all |r| < .05, all p > .40). A significant age difference in cognitive offloading emerged, with participants aged 36–45 reporting substantially lower AI reliance than the 17–25 cohort (p = .039). Gender did not significantly predict any outcome variable. The null correlation between cognitive offloading and critical thinking provides limited support for the AI-erosion hypothesis. Possible explanations include a ceiling effect in CT scores, strategic rather than habitual AI use, and the metacognitive sophistication of a pedagogically specialised sample. Findings carry implications for AI literacy integration in chemistry teacher education programmes.