Key points:
- AI could undermine the academic experience without checks and balances
- Building an everyday equitable practice for AI in schools
- Your institution’s AI strategy needs instructional designers
- For more news on AI in higher education, visit eCN’s AI in Education hub
AI, as they say, has entered the chat. The benefits of AI, when applied responsibly, can be a boon to the college and university environment, supporting instruction, study methods, accessibility, course design, administrative tasks and much more.
Its prevalence, however, creates new questions among faculty about content, research and even assessments: Did a student write this, or did AI? A recent student AI usage survey showed more than 63% of respondents use AI to support at least some of their coursework and more than a third say they use AI chatbots to help answer exam questions.
AI in higher education is not going away, but without checks and balances, regulation and usage guidelines, it threatens to undermine the value of the entire academic experience. When students stop learning, they lose the ability to think critically, solve problems independently, and advance their own knowledge in preparation for a post-college world. In addition, the value of the educational experience itself comes into question.
Proving the value of learning requires different ideas
A recent EDUCAUSE Horizon Report lists the pressures weighing on the academic sector, citing value, trust, transformation, AI adoption, cybersecurity, data privacy and accountability. A common denominator among these is academic integrity. As AI erodes integrity, higher education is being asked to prove that learning is real and leads to betterment. This, in turn, requires the ability to verify the conditions under which learning takes place and an ability to measure an increase in knowledge.
When it comes to AI and student misuse, many institutions lean on after-the-fact detection, but this approach is proving to be “too little, too late.” A better path is to initiate a verification process from the start, defining who completed the work, under what conditions, and what guardrails were in place that helped reduce “cheating” and supported measurable learning. With AI everywhere today, trust in assessment must be built into the entire learning process, not just evaluated after work is submitted.
The flaws of post-hoc AI detectors
AI detection tools certainly have a place in the academic integrity toolset. But an immediate problem with using AI detection (on its own) is timing. By the time an instructor investigates the source of a submission, both the instructor and the institution are already reacting to work completed and are actively seeking student misconduct. This creates an atmosphere of distrust among faculty and students alike, according to the EDUCAUSE report, and can negatively impact the learning environment.
Whether detection products produce accurate results is also under scrutiny. The EDUCAUSE Horizon Report illustrates a growing distrust in AI detection tools, citing concerns about accuracy, bias, false positives and procedural fairness. At the core of this distrust is the conflict that arises when students feel unjustly accused of wrongdoing. Studies show AI detectors struggle when generated text is edited by a person, paraphrased, translated, processed through another AI tool, or “humanized.” Unfortunately, AI detectors are more of a smoke detector than a provider of concrete evidence.
Researchers from the University of Florida found significant rates of false positives and false negatives, further confirming that existing AI detectors should not be used as the only basis for determining whether content was AI-generated. A University of Florida official pointed out their research “shows us that for as many studies as we see claiming that a certain percentage of academic work is AI-generated, we actually don’t have tools to measure any of that.”
What verification-first looks like in action
Moving from a model of scrutinizing content after submission to a verification-first process empowers faculty and supports those responsible for teaching and evaluating learning.
Leaders at Grand Canyon University decided to put the verification-first approach to the test. Its “verification-centered assessment framework” leans into faculty review and an emphasis on verifying student understanding without the reliance on AI detection software, according to the university. Faculty set strict guidelines for use of technology and create coursework and assessments that help students develop the decision-making skills necessary to use technology ethically and responsibly. It requires students to explain their thinking, apply concepts, and demonstrate mastery of course outcomes, and offers a relevant learning environment that mirrors what will be expected in professional environments.
This same model can be applied across all of higher education, and its flexibility means that universities can modify it as needed. At the core, however, recommended foundational components should include:
- Clear expectations for the use of technology resources and AI tools
- Confirmation of student identity, especially in testing scenarios
- Tap appropriate technology to limit access to unauthorized AI tools and other unauthorized materials.
- The use of human review and oversight throughout coursework
Beyond verification-first guidance, faculty can consider shifting how exams or assignments are created to reduce the ability for AI tools to be used. Instead of question-and-answer tests, ask students to review two case studies and respond to specific questions that require them to think critically and integrate their opinions and reasoning.
Work to link assignments directly to specific coursework materials or case discussions. Steer clear of broad, open-ended questions that AI can easily fill in with plausible answers and require submissions that include personal experience, local examples or unique scenarios that are unlikely to be found by AI.
When applied thoughtfully, the right tools can support both faculty and students. The right model is a beneficial balance: faculty have stronger context for academic judgment, while students complete assignments under clear and consistent expectations. That consistency is essential to preserving trust and integrity.
Embedding trust into learning
The influence of AI on how students learn, how faculty teach and how institutions operate will continue to evolve. Universities need clear, secure, and fair assessment conditions that preserve learning outcomes and support an environment of trust.
Post-hoc AI detection alone cannot carry that responsibility. Institutions need preventative processes that support students, empower faculty’s academic judgment and preserve confidence in the value of higher education.
Jordan Adair is VP of Product at Honorlock.
- Beyond the gotcha game: Moving from AI detection to verification-first supports stronger academic integrity - September 16, 2026
- Choose encouragement before negativity sets the tone - September 11, 2026
- Building an everyday equitable practice for AI in schools - September 9, 2026