Provenance Does Not Decide Authorship: What recorded AI involvement does and does not establish
Edward Meyman · Zenodo (CERN European Organization for Nuclear Research) · 2026
Suppose the question of AI involvement is settled by an authentic production record sufficiently complete for the events at issue, rather than by a detector estimate. What follows? Depending on its schema and capture scope, such a record may identify the participating tool, timing, and recorded inputs, outputs, and transformations. It does not, by itself, decide who directed the work intellectually, whether the assistance was permitted, whether disclosure was owed, or who answers for what the work asserts. The paper distinguishes intellectual contribution from intellectual directorship, defined as the exercise of judgment that controls the work's governing thesis, assertions, evidence, reasoning, revision, and final form, and pairs directorship with accountable adoption as a proposed criterion for institutional and normative authorship. It proposes an institutional and normative criterion, not a copyrightability test. Provenance is evidence about production. It is not an authorship verdict. A companion document, Sources and Verification Method, is deposited with this paper. It records where each factual claim comes from, how each source was obtained and verified, the evidence grade assigned to it, two positions the paper considered and rejected together with the evidence against them, and the limits of that record. This paper is the successor to Detection Is Not Provenance: Why an AI detector score is evidence about production, not a record of the production process (https://doi.org/10.5281/zenodo.21881429). The predecessor established that a detector estimate is evidence about production, not a record of the production process. This paper grants a reliable record and examines what conclusions may validly be drawn from it.