Abstract:

Multiple-choice questions (MCQs) are widely used in higher education due to their low cost and ability to measure learning outcomes. It is important to evaluate these tools and obtain quantitative feedback to continue improving the questions accordingly. This paper presents BankGen, an AI-driven framework that automates MCQ generation, management, and evaluation using generative AI. BankGen integrates Google Gemini with structured prompt engineering to generate curriculum-aligned questions mapped to course learning outcomes (CLOs), supports automatic exam creation, and performs item analysis using difficulty and discrimination indices. Implemented using Replit and evaluated by faculty members across disciplines, BankGen demonstrated strong curriculum relevance, CLO alignment, linguistic quality, and question clarity. Results indicate that BankGen can reduce assessment workload while supporting continuous improvement in assessment quality