This research explores the detection of AI-generated text, addressing challenges arising from its increasing prevalence in various domains such as journalism, education, and software development. While AI-generated text offers numerous applications, it also raises significant concerns about plagiarism, misinformation, and academic integrity. To address these concerns, this study employs advanced machine learning models, including BERT, DistilBERT, RoBERTa, and proposes a knowledge graph based BERT variant, for classification of human-written and AI-generated. A standard and publicly available dataset comprising 10,000 essays generated using predefined prompts was utilized for training and evaluation. The dataset includes a balanced mix of human-authored and AI-generated content, enabling the models to capture nuanced linguistic patterns. The methodology involves extensive preprocessing, tokenization, and fine-tuning of the models to optimize their classification performance. Among the models evaluated, the knowledge-integrated BERT demonstrated the highest accuracy, leveraging structured knowledge graphs to enhance contextual understanding. The study further evaluates the models using metrics such as accuracy, precision, recall, and F1-score, with results indicating a trade-off between computational efficiency and detection accuracy.