Abstract
Generative artificial intelligence can support scholarly writing, but fluent text does not establish evidentiary support, responsible authorship, or discoverability. Preprints, repositories, persistent identifiers, and author profiles introduce additional decisions that are often addressed separately from research integrity. This conceptual analysis proposes TRACE-AI, an auditable workflow connecting the evidentiary state of a research object to its preparation, disclosure, deposition, and discovery. A purposive documentary synthesis integrates published research with primary editorial and infrastructure documentation; it is not a systematic review or an evaluation of an implemented intervention. TRACE-AI comprises five stages: Type the object and specify its evidence state; use Responsible AI assistance with authorized inputs; Audit claims, methods, and references; Curate metadata, rights, identifiers, and versions; and Expose the object while separately verifying hosting, discovery, and attribution. Its central contribution is a traceability contract linking each consequential assertion to an identifiable source or analysis, an explicit scope, a recorded decision, and an accountable person. The framework distinguishes technical discoverability from epistemic justification: a deposited object, a resolving digital object identifier, or a profile entry cannot substitute for evidence supporting a claim. A hypothetical surgical-simulation protocol illustrates how an unsupported claim of effectiveness should be removed rather than polished. Reusable audit templates and candidate evaluation outcomes are specified, including unsupported-claim frequency, bibliographic accuracy, metadata completeness, correction traceability, and implementation burden. TRACE-AI has not undergone expert validation, reliability testing, or comparative effectiveness evaluation. It should therefore be treated as a testable conceptual and educational proposal, not as certification of research quality or a guarantee of indexing. The proposed workflow makes responsibilities and unresolved evidence visible before dissemination and provides a basis for prospective evaluation.