Abstract

This research introduces a conceptual and explainable framework that brings together Model-Driven Architecture (MDA), Model-View-Controller (MVC), Large Language Models (LLMs), and Human-in-the-Loop (HITL) methods to support intelligent web systems used in marketing. The framework was developed from recent studies on web intelligence, model-driven design, and explainable AI. It responds to three issues often discussed in this literature as system complexity, the need for transparency, and the role of stakeholder collaboration. Within the framework, business goals are organized through the MDA layers of CIM, PIM, and PSM, linked with MVC components, and paired with provenance tags (AI-only, AI + Human, Human-only) that strengthen accountability and interpretation. A small proof-of-concept exercise was carried out using real posts from the Thai-language Aizhongchina Facebook page, with ChatGPT-5 applied to explore how the framework could classify posts and assign confidence-based provenance labels for content-style classification in a content marketing context. The example helps show how structured modeling, explainable AI, and human review can work together while keeping brand tone and stakeholder oversight intact. The main contribution of the research is its combination of software-engineering logic with explainable and human-centered AI in one transparent framework. Although the work remains conceptual, it provides a practical model that can guide future implementation that future studies could test it on larger datasets, extend it to other marketing cases with real implementation, and include quantitative evaluation to confirm its performance.

Statistics