Over the past year, a growing list of universities has announced that it will stop using AI detection features, citing concerns about accuracy, fairness, and privacy. Some institutions kept their similarity-checking tools while retiring the AI-writing indicator, and some now say outright that detector output cannot be used on its own in misconduct cases. The shift reflects a recognition that detection is a weak foundation for integrity policy. Instructors who still need to protect the value of their writing assignments are looking for something sturdier.

The pressure on faculty is real. One university reported that integrity cases have risen roughly fourfold since the pandemic, and another saw a 170 percent increase over three academic years. Hearing committees are overwhelmed, and some are moving toward conversation-based models that resolve cases informally. In that environment, handling integrity through assignment design is both more humane and more efficient than relying on adjudication.

Higher education commentators increasingly describe detectors as audit inputs rather than enforcement tools, meaning they might prompt a closer look but never prove anything. The tools also lose ground as soon as a student paraphrases or edits the output. What institutions can actually control is the structure of assignments, the tools students are permitted to use, and the skills they are taught. Those levers work whether or not a detector is available.

Design assignments that make thinking visible

Break large papers into stages that each produce something observable. A proposal, an annotated source list, a first draft, a peer review response, and a final reflection on what changed make it difficult to outsource the whole project and easy to see how the author's thinking developed. Each stage can be graded lightly, with detailed attention reserved for the final essay. The trail itself becomes the primary evidence of authorship.

• Require staged submissions with brief, low-stakes feedback at each step.
• Tie prompts to specific class discussions, local data, or sources assigned this term.
• Add a short in-class or oral component that asks students to defend their central claim.
• Ask for a reflection on one revision decision and why it improved the argument.
• State in writing exactly which AI uses are allowed, with disclosure expected for any use.

Use in-person and oral checks selectively

Handwritten in-class essays and brief oral defenses have returned at many institutions, including programs that now run faculty workshops on oral assessment. These methods work because they require students to produce reasoning in real time. They are also costly, with oral exams in particular demanding substantial instructor time, and they can disadvantage students with anxiety, disabilities, or limited fluency. Using them selectively, for example as a ten-minute conversation about one paper, preserves much of the benefit.

A balanced approach pairs a take-home essay with a short in-person check on a sample of students rather than everyone. Students know any of them might be asked to explain their paper, which encourages genuine engagement, and instructors keep the workload manageable. The check should be framed as an ordinary part of the course, with a rubric and a clear purpose. That framing prevents it from feeling like a punishment.

Teach responsible AI use instead of pretending it away

Research on how students use AI suggests that most lean on it for quick answers, but faculty guidance increases the share who use it to clarify concepts and strengthen their own work. Spending a class session showing what acceptable use looks like, such as asking a chatbot to quiz you on a reading but not to write your thesis, shifts behavior more than a warning does. It also makes the line between help and substitution understandable. Students who know where that line sits are better able to stay on the right side of it.

Disclosure requirements add another layer. If students must note which tools they used and how, honest use becomes visible and acceptable, and dishonest use becomes a clear policy violation instead of a gray area. Several institutions report that disclosure-based policies reduce disputes because there is less to argue about. They also build habits students will need in workplaces that expect transparent AI use.

Keep grading consistent and focused on learning

A clear rubric ties the final grade to the quality of the argument, evidence, and organization, which lowers the stakes of authorship questions. When feedback names specific strengths and next steps for each criterion, students receive learning value even if their process was imperfect. Rubric-based grading tools that draft criterion-level comments for the instructor to review can keep that quality consistent across large sections without adding evenings of work. The instructor still makes every final decision.

Departments should also share what works. A brief meeting each term to compare assignment designs, rubrics, and integrity outcomes helps faculty avoid reinventing solutions and creates common expectations for students who take multiple courses. Over time, a shared approach to integrity that emphasizes design, disclosure, and feedback proves more durable than any single technology. Detectors may still play a minor auditing role, but they no longer need to carry the weight.