When ChatGPT Co-Authors: Rethinking Academic Integrity and Generative AI Policy in Higher Education

Abstract

When ChatGPT can produce a polished college essay in seconds, academic integrity faces a new challenge. Students are increasingly using generative AI to draft and revise their work, often without detection. The result is a structural mismatch between student behavior and institutional policy. Using Princeton University as a case study, this thesis draws on a systematic review of 461 course syllabi, institutional policy documents, student survey data, public online discussions, and evidence on AI detection tools. The findings show that AI use is widespread, while course-level policies are often absent or unclear. Nearly 60% of syllabi do not mention generative AI at all, leaving students to interpret expectations on their own. The problem is not simply a fragmented and inconsistent policy landscape, but that generative AI undermines the ability to enforce academic integrity altogether. Detection is unreliable, and the process behind student work is no longer visible, making it difficult to determine authorship or prove misuse. As a result, governance systems that rely on identifying and punishing violations begin to break down. Effective AI governance must move upstream, from individual instructors to the institution itself, by establishing clear institutional defaults, standardizing disclosure, and shifting evaluation toward the process behind student work rather than the final product. The central challenge is how universities can maintain the value and credibility of their degrees as these tools become embedded in how academic work is produced.