Beyond the Black Box: A Semantic Explainable AI Framework for Transparent and Trustworthy E-
Abstract
Artificial intelligence has become an important part of modern e-commerce, supporting product recommendations, search ranking, customer-review analysis, and personalized shopping experiences. Despite these benefits, customers often receive recommendations without understanding the reasons behind them. This lack of transparency can make it difficult to distinguish a genuinely relevant suggestion from one influenced by popularity, incomplete information, or commercial priorities. Explainable Artificial Intelligence (XAI) offers a way to make these decisions more understandable. However, explanations are useful only when they communicate relevant evidence accurately rather than simply presenting convincing language. This paper examines the relationship between XAI, semantic analysis, and transparent decision-making in e-commerce, with particular attention to research trends during 2024, 2025, and 2026. It proposes a conceptual framework that connects customer preferences, product attributes, review sentiment, recommendation models, and evidence-based explanations. The framework is designed to explain why a product may be relevant, which features support the recommendation, and where conflicting information remains. The paper also discusses explanation faithfulness, customer privacy, review reliability, fairness, and the limitations of language-model-generated explanations. Its contribution is a practical design approach for connecting recommendation decisions with verifiable semantic evidence. Future implementation and user testing are required to determine whether this approach improves customer understanding and decision quality in real shopping environments
