Using Generative Artificial Intelligence To Support Preparation For University Entrance Exams: A Systematic Review

Generative Artificial Intelligence (GAI) is transforming education, yet there is still a lack of systematization regarding its use in preparing for higher education entrance exams, such as the Brazilian National High School Exam (ENEM). This study aims to analyze, through a systematic literature review, the use of GAI as a support tool for exam preparation, focusing on questionnaire generation, study personalization, and error correction. A systematic review protocol was conducted across six databases, initially identifying 246 papers. After a three-stage selection process, nine primary studies were chosen for data extraction in order to answer five research questions. The results indicate a predominance of Large Language Models LLMs, particularly from the GPT family (RQ1), with accuracy being the main performance metric (RQ2). Questionnaire generation (RQ3) has evolved from simple prompts to Retrieval-Augmented Generation (RAG)). RAG was identified as the main technical strategy to mitigate “hallucinations” (RQ4), a core risk of this technology. Natural Language Processing (RQ5) underpins all applications, enabling everything from automated assessment to personalized tutoring. It is concluded that GAI is a powerful tool for exam preparation; however, its effective use depends on strategies such as RAG and, crucially, on human curation and validation. The main limitations of this mapping include the restricted search string focused on the term “ENEM” and the inherent risks of the technology, such as academic dishonesty and hallucinations.