Dyslexia is a neurodevelopmental learning disorder that significantly affects children’s reading and writing skills despite normal intelligence, and delayed identification may lead to long-term academic and psychosocial consequences. Existing dyslexia screening methods rely heavily on expert-driven assessments that are time-consuming, subjective, and difficult to scale in non-clinical settings. Although recent studies have explored artificial intelligence (AI) approaches for dyslexia detection, many remain limited to single-modality data, offline analysis, or non-mobile implementations, restricting their practical applicability for early screening. This study aimed to develop an AI-based mobile application for early dyslexia detection by leveraging sequential text and speech data through a Recurrent Neural Network (RNN) architecture, specifically the Gated Recurrent Unit (GRU). A Research and Development (R&D) methodology was employed, encompassing requirements analysis, system design, GRU model training, mobile application development with Flutter, and system integration with a RESTful backend and a MySQL database. The GRU model was trained on preprocessed reading text and voice recordings to capture temporal patterns associated with dyslexia-related reading behaviors. Experimental results indicate that the proposed model achieved reliable classification performance in identifying dyslexia-related patterns, while the mobile application successfully delivered real-time screening results and maintained longitudinal assessment records. The findings demonstrate that integrating lightweight sequential deep learning models into mobile platforms offers a scalable and accessible solution for early dyslexia screening, supporting independent use by parents and educators outside clinical environments.