Construction and Case Study of a Quality Evaluation Model for Intelligent Information Systems in a Generative Artificial Intelligence Environment: The Case of Bilibili

Abstract

After generative artificial intelligence (AI) was embedded in search, content generation, and user interaction, traditional information system quality evaluation faced new boundaries. Based on the DeLone and McLean (D&M) Information Systems Success Model, this study adopts literature research and a single-case study, takes Bilibili as the case, and constructs and applies a six-dimensional evaluation framework based on public materials, observations of platform functions, and publicly available user feedback. The findings show that system quality, information quality, and service quality can still explain basic platform operation, information output, and service support, but they cannot fully cover AI’s semantic understanding and task completion, the trustworthiness of generated results, and continuous human-computer interaction. Therefore, this study introduces AI intelligence quality, trustworthiness quality, and interaction quality on the basis of the three traditional dimensions and clarifies the boundary of each dimension. The Bilibili case shows that the framework can identify both basic capabilities and AI-specific quality, thereby helping to avoid conflating technical performance, content quality, intelligent capability, and trust issues. The study provides a structured analytical framework for evaluating the quality of intelligent information systems in a generative AI environment and a reference for platforms to optimize AI functions and governance mechanisms.