AI and critical thinking: cognitive partner or cognitive substitute

Artificial intelligence is reshaping higher education faster than pedagogy can respond. AI tools can personalise feedback, scaffold metacognition, and deepen collaborative inquiry, yet without structured guidance, they equally risk creating cognitive dependency, distorted self-assessment, and superficial thinking. Educators urgently need a coherent framework to navigate this tension. This paper draws on a systematic review of 63 peer-reviewed studies (January 2023 April 2025, PRISMA 2020) to pursue three objectives: synthesising evidence on AI's dual impact on critical thinking; examining ethical challenges including academic integrity, algorithmic bias, and privacy; and proposing the AI-Enhanced Critical Thinking (AI-CTE) Framework. Unlike existing AI literacy models, the AI-CTE framework explicitly positions critical thinking preservation rather than AI proficiency as the central pedagogical objective. The evidence reveals that AI functions as either a cognitive partner or a cognitive substitute the difference lies entirely in pedagogical intentionality. The AI-CTE Framework addresses this through four progressive dimensions: AI Literacy and Foundational Understanding, Guided Application and Critical Engagement, Evaluation Synthesis and Creative Construction, and Ethical Reflection and Responsible Practice, united by a feedback loop that keeps the learner as active cognitive agent. Empirical validation across diverse educational contexts remains a priority