AI Has Changed Academic Integrity Forever

Diversified Education Service’s Aron Boxer offers this commentary on how AI has changed academic integrity forever. This article originally appeared in Insight Jam, an enterprise IT community that enables human conversation on AI.

Catching cheaters won’t fix it. Ever since ChatGPT went online in late 2022, it hasn’t taken long for students to recognize that their academic future has permanently changed. Teachers acknowledge it too, but at the beginning of the modern AI era, they approached this revolutionary change in full denial. Nearly four years later, nobody has figured out what to actually do about it.

The “problem” is hard to argue with at scale. A February 2026 College Board survey of more than 3,000 U.S. college faculty members found that 74 percent report students using AI to write essays or papers, and 67 percent say students use it to paraphrase or rewrite content.  Also, a whopping 92 percent are concerned about AI-driven dishonesty.

A Real-Life Example

As an executive function coach, I work with a 10th-grade student who disregarded a paper he and I had drafted together and instead used ChatGPT to complete it. Upon reviewing the completed assignment, the absence of human authorship made it clear that automated detection software was unnecessary. The teacher was justified in giving him a zero, but since the paper was a graduation requirement for all Connecticut students, he had to rewrite and pass. Consequently, the student was made to handwrite the assignment under supervision, and he received a 60, the minimum passing grade.

Where it gets tricky is that the teacher left the zero on the original, effectively giving the student a 30 percent on the primary quarter assignment, tanking the student’s grade into the D-range. Whether or not you agree with the teacher’s approach, such punitive measures are a last-ditch effort to control a technology that advances far faster than human ability.

But here’s where it gets complicated: the tools we have to detect AI-generated work are unreliable.

AI Detection Tools are Unreliable

Turnitin is the most widely used plagiarism detection platform, and it now reports an AI content score using a similar rating system. AI detectors are unreliable, and false positives are quite common.

Dr. Torrey Trust, a professor of technology at UMass Amherst, has made the point plainly: there’s no reliable way to determine if a student’s work is AI-generated. Detectors are notoriously unreliable.

As a value add with my students, I use Winston.ai, and through trial and error have found it to be the most reliable. However, I concur with Dr. Trust’s assessment. They are unreliable in the same way lie detector analysis is not admissible in a court of law.

“AI-Assisted” vs. “AI-Generated” 

Beyond detection, the line between “used AI” and “cheated” has become nearly impossible to determine.

Students ask ChatGPT to write whole essays. They use Grammarly to restructure sentences, Copilot to autocomplete paragraphs, and Gemini in Google Docs to rework entire essays. Now, AI is embedded in almost every aspect of the work students already do. Trying to separate what’s “AI-generated” from what’s “human” is something no one—neither ethicists nor university administrators—has convincingly managed to do.

The Case for Teaching AI Literacy Instead of Punitive Measures

My contention is that if educators devoted as much energy to teaching AI literacy as they do to policing it, the education system wouldn’t be underwater in cheating. Rather than policing student AI use, institutions should embed AI literacy into the curriculum. That includes developing judgment about the ethical use of AI.

Math was once viewed as the end of math as we knew it. But did schools revert to the abacus? No, they embraced the technology, and now most standardized tests allow its use.

The bottom line is not the ethical misuse of AI—it’s the ever-growing scale of “uncritical thinking.” The consequences are well beyond a campus. It will follow younger generations into the workplace, where ironically, they are using a technology that will render them obsolete. At this point, Gen Z and Alpha are effectively training their replacement and diminishing their relevancy.

Better detection software won’t solve that. What might fix the issue is readdressing what we’re actually trying to assess. We need to give both students and faculty the frameworks to engage honestly with AI, rather than just finding smarter ways to hide that they’re using it at all. Because what will ultimately be hidden is human relevance in the market economy.