From Algorithms to Tutors: Tracing the Evolution of Generative AI in Education

The integration of Artificial Intelligence (AI) into educational settings has historically been characterized by reactive, rule-based systems designed for administrative efficiency and corrective feedback. Early technologies, relying on decision trees and support vector machines, functioned primarily as digital gatekeepers—flagging errors without offering the reasoning capabilities required for deep learning. However, the recent emergence of Generative AI (GenAI) and Large Language Models (LLMs) marks a significant evolutionary leap, shifting the pedagogical paradigm from computer-assisted instruction to AI-driven cognitive partnership. This paper traces the evolutionary trajectory of AI in education through three distinct phases: (1) the Era of Automation, defined by rigid Intelligent Tutoring Systems that struggled with unstructured domains; (2) the Era of Augmentation, characterized by predictive analytics and early warning systems; and (3) the current Era of Agency, where generative models act as co-creators and Socratic tutors. Through a PRISMA-guided systematic review of literature from 1984 to 2025 (n = 38), this study analyses how GenAI addresses Benjamin Bloom’s "2 Sigma Problem" by providing scalable, personalized tutoring, while simultaneously introducing critical risks such as hallucination and cognitive offloading. The findings indicate that as AI evolves from a tool for correction to a tool for creation, traditional output-based assessment becomes obsolete, necessitating a pivot toward process-based assessment and critical AI literacy. The paper concludes with a forecast for 2030, predicting a transition toward multimodal, autonomous agentic systems and sovereign educational models that will redefine the teacher’s role from a distributor of content to an architect of cognitive inquiry.