Automatic Text Summarization (ATS) involves condensing long texts into concise summaries while retaining the core meaning, addressing the challenge of processing vast amounts of textual data. Traditional methods often lack the precision and efficiency required for such tasks, which requires the development of scalable result using modern Natural Language Processing models. This study proposed a deep learning-based ATS system, leveraging advanced models such as Text-to-Text Transfer Transformer (T5-small) and T5-base, Generative Pre-trained Transformers-2 (GPT-2), and Longformer Encoder-Decoder (LED) to generate summaries from a specialized psychological dataset. The research compares the performance of all the employed models to get the better model which has high Recall-Oriented Understudy for Gisting Evaluation (ROUGE) score. Extensive preprocessing and feature extraction techniques were employed to prepare the dataset, followed by model training and evaluation using the ROUGE metric. The LED model achieved the highest ROUGE scores of Rouge-1: 0. 6269, Rouge-2: 0. 3921, Rouge-L: 0. 5261, and Rouge-Lsum: 0. 5246, indicating superior performance in producing accurate and coherent summaries than other employed models. The LED model has been deployed via a mobile and web application using FastAPI, offering users an efficient tool for summarization and FAQs. This work demonstrates the potential of deep learning models in enhancing the quality and accuracy of text summarization, providing a robust solution for handling textual data.