Provenance-Aware Hybrid Search in AI-Mediated Digital Libraries

Abstract

Digital library search has traditionally supported source-centered discovery by enabling users to inspect records, metadata, citations, and related resources. Recent advances in retrieval-augmented generation (RAG) and AI-mediated search are introducing answer-oriented forms of interaction, in which synthesized responses may appear before users engage with the underlying source records. This raises the risk of zero-click retrieval, where generated answers may become endpoints rather than pathways into digital library collections. This paper proposes a provenance-aware hybrid search framework for digital libraries. Using a conceptual and design-oriented synthesis of prior work on hybrid retrieval, RAG, provenance, explainable information retrieval, and direct-answer search, the proposed framework connects digital library content and metadata with hybrid retrieval and AI synthesis, provenance-aware interface design, and user interaction. It identifies five user-facing design affordances: source visibility, metadata context, citation traceability, exploratory browsing, and trust-supporting provenance cues. By positioning generated answers as entry points into traceable source networks, the framework provides a conceptual model for maintaining source visibility and exploratory access in AI-mediated digital library search.