Nowadays, the AI-generated texts are more common to encounter in different contexts, and it is a hard task to determine when a text is human or AI generated. This sets an ethical problem in the academic community. Here we present a study that aims to develop a highly accurate way to determine when a text is AI-generated or human-generated. We propose a model that integrates a Long-Short Term Memory Network in addition with topological data analysis for detecting AI-generated texts. We will use topological descriptors like persistent homology to capture essential features of the text and apply persistent images to distinguish between human and AI-generated texts. Additionally, our proposed model has the capability to accurately classify question-answer texts with a small training dataset.