Parkinson’s Disease (PD) is a neurodegenerative condition that progressively affects both motor and non-motor functions, profoundly affecting the daily lives of those it touches. Timely detection and precise evaluation of the disease’s severity are essential not only for optimizing treatment strategies but also for empowering patients and their families to manage this complex condition more effectively. This study introduces a novel approach that employs an EEG-driven deep Convolutional Neural Network (CNN) model to assess the severity of PD. We analyzed electroencephalography (EEG) recordings from 20 PD patients and 20 healthy individuals to trained a deep learning model aimed at classifying the levels of disease severity based on the Movement Disorder Society Unified PD Rating Scale (MDS-UPDRS) score. The model demonstrated outstanding performance, particularly in the eyes-open condition, achieving an accuracy of 99.46%. These results underscore the model’s impressive ability to distinguish between healthy controls and individuals with varying severity levels of PD (mildly affected and affected). Although further validation with larger clinical samples is required, these findings suggest that integrating EEG signals with deep learning techniques could be a highly effective approach for the early diagnosis and management of PD.