The Wrong Battle: Why Your Institution's AI Policy Is Probably Solving the Wrong Problem

• By B. Jean Mandernach, Ph.D.
• 05/28/26

Every week, faculty members across higher education are spending hours doing the same thing: trying to figure out whether a student actually wrote a paper. They're running submissions through AI detectors. They're Googling suspicious phrases. They're comparing sentence-level complexity across a student's body of work. And they're losing.

Not because they aren't smart or dedicated. They're losing because they're fighting the wrong battle.

The conversation on most campuses has become consumed with detection: How do we catch students using AI when they shouldn't? The impulse to protect academic integrity is legitimate, but the detection-first approach has a fatal flaw. AI detectors regularly flag legitimate student writing as AI-generated, including work by students who used only grammar tools, while missing AI-generated content that has been lightly edited. The bias problem compounds the accuracy problem: Stanford researchers found that detectors misclassified over 61% of essays written by non-native English speakers as AI-generated. A 2023 study in the International Journal for Educational Integrity that tested 14 detection tools concluded they are neither accurate nor reliable. As Bowen and Watson have argued, the question institutions must honestly confront is how many false accusations they are willing to accept as collateral damage. The tools students are using are evolving faster than any institution can keep pace with, and the arms race is unwinnable. In the meantime, institutions are spending enormous energy on policing rather than teaching.

There's a deeper problem with this framing, though, and it's one that gets far less attention. Focusing on detection treats the symptom, not the disease. The real challenge isn't that students are using AI. It's that AI use has fundamentally undermined the validity of many assessment tools that higher education has relied on for decades. A five-paragraph essay, an end-of-semester research paper, a take-home case study: These were always proxies for learning, never the learning itself. AI hasn't changed that. It has just made the gap between the proxy and the thing it's supposed to measure impossible to ignore.

That realization is the beginning of a genuine institutional response.

The Paradigm Shift Administrators Must Lead

Institutions that are navigating this well aren't asking, "How do we catch students using AI?" They're asking a different question entirely: "How do we know if our students are actually learning?"

That shift in question changes everything downstream: policy, assessment design, faculty development, and institutional culture. And it requires leadership. Faculty can't make this pivot in isolation. The framing has to come from the top, because what's really being asked of faculty is a significant professional and intellectual reorientation.

At Grand Canyon University, our approach rests on three interconnected pillars: a clear institutional position, curricular modernization, and what we call learning integrity, a framework that empowers faculty to verify learning rather than detect misconduct.

The policy comes first, but not for the reasons you might expect. A clear, published institutional policy on AI use isn't primarily a disciplinary document. It's a communication tool. It signals to faculty, students, and the broader academic community what the institution believes about AI, what it expects from everyone, and what the institution is prepared to support. Without that signal from leadership, faculty are left to improvise, and improvisation tends to default toward detection because detection feels like doing something.

What the Policy Has to Get Right

An effective AI policy does three things that detection-focused policies typically don't.

First, it acknowledges that AI use exists on a spectrum. Submitting a fully AI-written essay as your own work is categorically different from using AI to brainstorm, check grammar, or develop a first draft you then substantially revise. Policies that treat all AI use as equivalent create ambiguity that students navigate poorly and faculty enforce inconsistently. The policy needs to be clear about what constitutes appropriate use, what requires disclosure, and what is genuinely prohibited, and those distinctions should be tied to learning objectives rather than to a blanket prohibition on AI tools.

Second, it shifts accountability from AI use to learning demonstration. Students should understand from the outset that what they will ultimately be held responsible for is demonstrating their own understanding. The question isn't whether you used AI; it's whether you can show you've learned. That framing changes the stakes for students and transforms the relationship between students and faculty from adversarial to educationally purposeful.

Third, it empowers faculty explicitly. When a faculty member has genuine questions about whether a student's submitted work reflects their actual understanding, the policy should give them clear authority and clear tools to find out. This is what we call learning verification, and it is the operational core of a learning integrity framework.

Learning Verification, Not Detection

Verification is not investigation. The distinction matters enormously, and administrators need to be precise about it when communicating with faculty.

Investigation asks: Did this student cheat? Verification asks: Does this student understand what they submitted? The first question puts the faculty member in an adversarial, prosecutorial role and requires a standard of proof that is essentially impossible to meet. The second question puts the faculty member in their natural role as an educator assessing learning, which is entirely within their professional authority.

Verification can take many forms: a brief conversation about the ideas in a submission, a short follow-up question about methodology, an oral explanation of a process. The point is not to catch a student doing something wrong; it's to confirm that the learning objectives of the assignment were met. If they were, the student passes. If a student cannot demonstrate basic understanding of work they submitted, that is itself evidence about their learning, evidence that should affect the grade not as punishment, but as an accurate reflection of what was actually learned.

This reframing matters for faculty culture as much as it matters for practice. One of the most significant challenges in implementing this approach is that faculty are deeply reluctant to let go of the detection mindset. It feels, to many of them, like giving up, like conceding the integrity of their courses. What they come to understand, over time, is that a learning verification conversation is actually a more rigorous and more defensible assessment than a plagiarism detector score. It puts the educational relationship back at the center of the integrity question, where it belongs.

Curricular Modernization Has to Happen in Parallel

Verification handles the immediate assessment challenge, but it doesn't solve the underlying structural problem. If your assessments are still primarily long-form written assignments that students complete without supervision, you are designing for a pre-AI world. Institutions need to modernize how they assess learning alongside changing how they respond to AI use, not as a separate initiative to be addressed later.

That doesn't mean eliminating writing. It means adding performance-based tasks, oral presentations, process documentation, in-class demonstrations, and sequenced assignments that make the development of thinking visible. It means designing assessments where AI assistance, even if present, is part of a visible process the student has to understand and explain. It means asking faculty to think deliberately about which assignments are designed to assess learning, and whether those assignments would still function as valid measures of understanding if a student had AI assistance.

This is not a small ask. It requires curriculum review, faculty development, and time. But it is the work that actually addresses the problem, rather than the symptom.

What Leadership Actually Looks Like

The hardest part of leading this shift is not writing the policy. It's changing the faculty culture. That requires administrators to do something counterintuitive: give faculty permission to stop fighting the detection battle, and make clear that doing so is not a failure of academic standards but an evolution of them.

In practice, this means professional development that goes beyond policy training to a genuine reexamination of what assessment is for. It means creating space for faculty to share and learn from each other's verification practices. It means being honest with faculty that AI detectors cannot be the foundation of an academic integrity system, and that the institution will not ask them to rely on tools that are demonstrably unreliable.

It also means being clear about what the institution will back faculty on. A faculty member who conducts a verification conversation and determines, based on that conversation, that a student cannot demonstrate understanding of their own submission needs to know the institution will support their professional judgment. Policy without institutional backing is hollow.

The question students are implicitly asking when they reach for AI to complete an assignment is whether the assignment is actually about learning or about production. Institutions that answer that question seriously, with redesigned assessments and a genuine commitment to learning integrity, will find that the AI problem becomes significantly more manageable. Not because AI goes away, but because the educational experience is designed so that AI use doesn't short-circuit the learning it's meant to produce.

The goal was never to catch students. The goal was always to teach them. It's time to build systems that reflect that.

AI Disclosure: Claude was used as a writing support tool to assist with drafting, organization, and refinement of language. All ideas, interpretations, reflections, judgments, and recommendations presented are the original work of the author. The AI functioned as a collaborative writing partner under the author's direction, with all substantive decisions and content ownership remaining with the author.