Leticia Vieira Gonçalves, Anderson Mateus de Souza, Samara Martins Nascimento Gonçalves, Reudismam Rolim de Sousa, Verônica Maria Lima Silva
The Generative Artificial Intelligence (GAI) in the educational context raises the need for a systematic mapping of its applications and implications. This work aims to conduct a comparative analysis of the performance of four GAI tools for knowledge discovery, such as ChatGPT (OpenAI), Gemini (Google), NotebookLM (Google), and DeepSeek (DeepSeek AI). The methodology consists of a comparative case use analysis, of a qualitative and exploratory-descriptive nature, based on the standardized execution of five pedagogical tasks (deep research for lesson planning, creation of slide structures, image generation, quiz elaboration, and podcast production) in each tool, totaling 20 analyzed interactions. The results demonstrate that the tools have distinct vocations: Gemini stood out for its native multimodality and resources for research and active learning (Canvas, Deep Research); ChatGPT established itself as an intellectual partner for writing and textual refinement; NotebookLM showed high reliability for academic research, with source citation; and DeepSeek presented good performance in text generation but with multimodal limitations. It is concluded that there is no single ideal tool, but an ecosystem of assistants with complementary specializations, whose effective use requires pedagogical mediation and teacher training for an ethical and critical integration into the educational process.