Credibility displacement in algorithmic financial reporting: a critical review of AI, disclosure credibility and capital-market consequences

Purpose This review examines how financial-reporting credibility is shaped when algorithmic systems enter the reporting chain. It asks where credibility is produced, tested and understood when artificial intelligence (AI), machine learning, large language models, eXtensible Business Reporting Language (XBRL) analytics and related tools affect reporting information. Design/methodology/approach The paper combines a search and screening procedure informed by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) with a concept-centric synthesis. The evidence base comprises a 226-paper curated core and a 144-paper close-synthesis subset. Sensitivity analyses re-examine the 39-record outlet boundary and the close-synthesis threshold. Findings The evidence remains dispersed across research on textual disclosure and natural language processing (NLP), audit analytics and assurance, structured reporting using XBRL, market interpretation, accounting information systems (AIS) implementation and AI governance. Across these streams, algorithms can improve reporting processes but may also alter the grounds on which reporting assertions are accepted. We describe the latter possibility as credibility displacement: model-related output, provenance or validation partly substitutes for scrutiny of the underlying assertion or for identifiable professional and institutional safeguards. Originality/value The review defines algorithmic financial reporting as a reporting-chain phenomenon rather than a general label for technology adoption. It distinguishes four sites: preparation, assurance, machine-readable presentation and algorithmic interpretation. Governance operates across these sites as a cross-cutting condition, making the proposed displacement mechanism more focused and empirically testable.