AI Co-Authorship: Why We Should Move Beyond Plagiarism and Academic Misconduct

The rapid integration of large language models (LLMs) into academic work has intensified concerns around plagiarism and academic misconduct. These concerns, however, rest on assumptions about authorship and contribution that no longer reflect how scholarly knowledge is produced. In conventional academic practice, multi-authored papers are routinely published without precise attribution of individual contributions, and this opacity is institutionally accepted. Human–AI collaboration often exhibits the same characteristics: the interaction is iterative, dynamic, topic-dependent, and frequently produces outcomes that cannot be cleanly partitioned between contributors—even by the human author. This paper argues that treating such collaboration as a form of co-authorship is both consistent with existing authorship norms and more intellectually honest than current disclosure-based approaches. Rather than focusing on tracing textual origins or policing production processes, scholarly evaluation should prioritize the quality, originality, and usefulness of the resulting work. The paper advances this argument through four interlocking principles—consistency, inherent opacity, misplaced evaluation, and value-primacy—and addresses the strongest objections, including the accountability gap and the risk of quality flooding. Moving beyond plagiarism and academic misconduct as primary evaluative frameworks allows academic institutions to better align ethical standards with contemporary practices of knowledge production.