Despite the fundamental importance of graphical models across numerous disciplines, educators face significant challenges in teaching advanced concepts in an understandable manner. This paper pro-vides a comprehensive review of 129 studies that have investigated automated assessment systems in graphical modeling education. The review is organized around seven research questions that address the critical needs and challenges of intelligent tutoring systems in this area. The analysis reveals per-sistent shortcomings – such as limited adaptability, fragmented methodologies, and a surprising ab-sence of Large Language Model (LLM) applications – despite their growing importance. Drawing up-on effective approaches identified in the reviewed literature, we offer a set of recommendations that could inform the design of a more dynamic, AI-driven tutoring environment. By incorporating modern technologies such as GPT-4, these recommendations aim to provide adaptive guidance, real-time sup-port, and advanced semantic analysis, thereby overcoming current limitations.