Academic Integrity in the Context of AI

The proliferation of generative AI technologies has complicated the state of academic integrity in higher education, given the blurring lines between novel and assisted output generation. Below, we outline what we mean by academic integrity and academic misconduct, research on the extent of academic misconduct in the context of AI, and what instructors can do to promote learning, trust, and ethical use of AI.Ìý

What Is Academic Integrity?

Academic Integrity refers to a commitment to the core values of honesty, trust, fairness, respect, responsibility, and courage (). These values enable:Ìý

  • Consistent and fair evaluation of student work
  • Appropriate recognition for student effort and learning
  • Credibility and trust in the learning process

Academic integrity is therefore not just about rule compliance, but also about cultivating ethical reasoning and action.

What Is Academic Misconduct?

Academic misconduct refers to any action that violates principles of academic integrity. Â鶹Ãâ·Ñ°æÏÂÔØBoulder’sÌýhonor code considers any behavior that could result in an unfair academic advantage as academic misconduct. Some examples include:

  • Using unauthorized materials or tools (e.g., notes, study aids, generative AI tools)Ìý
  • Portraying another’s work as one’s ownÌý
  • Aiding misconduct by sharing homework with peers or uploading course materials on third-party sites without the instructor's permission.

You can review thisÌýshort primer on Â鶹Ãâ·Ñ°æÏÂÔØBoulder’s Honor codeÌý

Is AI Increasing Academic Misconduct? What the Evidence SaysÌý

Academic misconduct is not a new problem. Research across multiple decades shows that misconduct rates have consistently been higher than 45%, sometimes higher than 88% ().Ìý

Disruptions to assessment practices can temporarily change rates of academic misconduct

COVID-19-related shifts: Increased adoption of remote or asynchronous exams has been associated with academic misconduct, often exceeding 54%. Nonetheless, these estimates are within the range of previously reported rates of academic misconduct. ().Ìý

AI-related shifts:ÌýSince the release of ChatGPT, some studies report unauthorized use of AI to have increased fourfold, with up to 45% of students reporting using AI in ways that are explicitly prohibited by course policies (;Ìý;Ìý;Ìý).Ìý

However, multiple studies indicate that the overall rates of misconduct have remained the same as those predating generative AI tools (;Ìý;Ìý;Ìý).Ìý

Why Does Increased AI Use Not Always Translate to Misconduct?

Several factors contribute to the gap between the increased use of AI in coursework and reported rates of academic misconduct:Ìý

  • Instructor overestimation:ÌýDistinguishing between human-generated and AI-generated text is notoriously difficult.ÌýIncreased scrutiny and uncertainty around AI use can amplify perceptions of misconduct without corresponding findings of misconduct (;Ìý)
  • Limits of detection tools: Moreover, AI detectors are unreliable and introduce false positives. They are also heavily biased against non-native speakers of English (;Ìý;Ìý). As a result, many suspected cases are possibly dismissed for lack of evidence.
  • Student underreporting:ÌýStudents may underestimate or justify the extent to which their AI use is reasonable. So even if used inappropriately, they are unlikely to cite or admit AI use if it is perceived as being socially undesirable (;Ìý)
  • How students actually use AI:ÌýAlthough over 88% students of students report using AI tools, most use them for initial ideation, troubleshooting, or when they get stuck—rather than complete entire sections of academic work (ETRA, 2026;Ìý)
  • Traditional forms of academic misconduct have declined:ÌýStudies have shown that students now seem to rely on AI to complete assessments instead of other types of academic misconduct, such as plagiarism from other sources, contract cheating, or copying work of peers ()

What This Means for Instructors

The above evidence suggests that AI may be reshaping how students seek support, rather than the rates of misconduct itself. Research shows that academic misconduct declines with increased institutional and instructional support (). However, given the ubiquity of AI tools, surveillance or prohibition alone is unlikely to work. Instead, effective approaches should focus on addressing:Ìý

Recommended Resources: