Design and Implementation of an AI Chatbot for Prospective Student Support

Abstract

Traditional communication channels in higher education often struggle to handle the high volume of admission-related inquiries, leading to delays and limited access to accurate information. This study presents the design, implementation, and preliminary evaluation of an AI-powered chatbot developed to support prospective undergraduate students at Tai Solarin University of Education, Ijebu-Ode. The system integrates an institution-specific knowledge base with a Large Language Model (LLM) to respond to queries on admission requirements, application procedures, available programs, and general university information. The chatbot was implemented via a web-based interface with Firebase backend services and Google-based authentication, with natural language responses generated using the Google Gemini API. Evaluation was conducted using 20 admission-related queries derived from real-world student information needs. Responses were assessed by the researcher and two evaluators based on accuracy and relevance, using official university information as a reference. Results show that 19 out of 20 responses were fully accurate and one was partially accurate, reflecting strong performance within the defined scope. The system demonstrated consistent relevance across all queries, with response times ranging from 2 to 10 seconds depending on network conditions. Limitations include the small dataset and qualitative assessment approach. These findings establish the chatbot as an effective proof of concept for structured admission inquiries. Further research involving larger datasets and quantitative metrics is needed to validate scalability and real-world effectiveness.