Multi-Class AI Text Detection Using XAI

It has become increasingly difficult to recognize whether a piece of writing was created by a human or generated by an AI model. As language models grow more powerful, the gap between machine-generated and human-written text is narrowing fast. This project explores a basic, interpretable machine learning setup that attempts to separate the two. We use a TF-IDF-based feature extraction method along with logistic regression to classify input text. To give more insight into how the system works, we integrate LIME, a tool that explains which words played a role in the final decision.We also built a simple web interface using Flask that lets users try the model in real time by entering text, seeing predictions, viewing explanations, and checking recent results.The model was trained on a small, balanced dataset that includes both human-written and AI-generated text.The model was trained on a small dataset with a balanced mix of human and AI-written text.It’s not perfect, but it gets the job done and isn’t hard to use. People like teachers, editors, or anyone curious about where a piece of writing came from might find it useful. Later on, we hope to add more training data and try better models. Even then, we want to keep things simple so anyone can use it without much effort.