Rethinking assessment integrity in the GenAI era: Making provenance the design principle

Higher education providers worldwide have responded to student use of generative artificial intelligence (GenAI) largely through post-hoc detection, but authorship verification does not address the validity problem created when assignments no longer require the disciplinary reasoning they claim to assess. This article advances a provenance-first alternative and specifies Process-Embedded Assessment with Verifiable Provenance (PEAVP). PEAVP relocates evidence from polished products to inspectable processes by coupling: a concise, structured provenance statement that bounds and discloses GenAI’s role relative to intended learning outcomes; a small learner dossier that makes key epistemic moves visible and gradable; and a brief structured defence that tests ownership and transfer. We argue that PEAVP restores construct representation, reduces construct-irrelevant variance from hidden GenAI mediation, and is administrable with existing tools.