Zero-click AI epistemic injustice and the governance of digital knowledge infrastructures

This paper examines zero-click AI, systems that deliver synthesized answers directly within platform interfaces without prominently visible source-level provenance. Such designs restructure conditions of credibility, visibility, and authorship by displacing upstream creators, suppressing provenance, and concentrating authority. I argue that these systems position platforms as functional epistemic agents whose design choices reliably generate epistemic harms. To analyze this, I develop a three-layer framework, focusing on infrastructures, appropriation, and governance, that specifies the mechanisms by which epistemic injustices arise under prevailing configurations. Building on debates in epistemic injustice theory, I show how zero-click systems systematically efface attribution, marginalize plural knowledges, and entrench epistemologies of ignorance. The normative argument is conditional rather than metaphysical: harms are contingent on design and policy logics, not inevitable features of automation. On this basis, I propose reform thresholds, attribution by default, provenance affordances, redistributive mechanisms, and pluralistic sourcing, derived from principles of proportionality and value-sensitive design. Properly operationalized, these measures demonstrate how digital infrastructures might preserve accessibility and efficiency while sustaining the pluralism required for a resilient knowledge commons.