Teachers are redesigning classes to prove learning, not catch AI cheats

A growing number of educators are abandoning the cat-and-mouse game of AI detection and instead redesigning their classes to make students prove they are learning. Rachael Zeleny, director of the writing program at the University of Baltimore, now recommends her instructors use a free app called Brisk to flag potential chatbot-authored work so they can skip grading it and focus on students' actual understanding Washington Post. The approach reflects a broader shift in AI in education assessment: rather than relying on flawed detectors or reverting to pen-and-paper assignments, educators are rethinking the structure of assessments themselves.

Why teachers are moving beyond AI detection

Traditional AI detectors have well-known accuracy problems, and the arms race between detection tools and increasingly sophisticated chatbots has left many instructors frustrated. The Laguna Beach Unified School District, working with Stanford researchers, found that a culture of suspicion had emerged where teachers spent energy catching cheaters while students felt guilty for using AI at all The 74 Million. In response, instructors are shifting from trying to catch AI-generated essays to designing assignments that require students to show their thinking process, engage in structured tasks, and participate in checkpoints that are harder to automate. This includes bringing back oral exams and verbal defenses, which are already being mandated for high school students in Denmark CNN.

What this means for AI builders and edtech products

For product teams building educational tools, this trend points to a clear opportunity. The demand is moving away from detection-centric products toward tools that help instructors evaluate learning processes. Brisk is one example: a free app that flags likely AI-generated text so teachers can triage their grading time, but it doesn't claim to be a conclusive detector. The real value is in scaffolding assignments that capture a student's reasoning, sources, and revision history. Builders who can create lightweight tools for in-class observation, structured peer reviews, or oral defense workflows will find eager buyers among schools and universities looking for scalable, fair assessment methods Washington Post. The shift also affects how institutions think about plagiarism policies. Many states have issued AI guidance for schools, but most lack concrete strategies for preventing or detecting cheating, leaving teachers to improvise Chalkbeat. That gap is a signal for edtech solutions that combine assessment design with integrity.

Practical trade-offs and limitations

The move toward learning-proof assessments has clear appeal, but it's not a silver bullet. Oral defenses and process-based assignments require more instructor time and training. Scaling these methods across large lecture courses or underfunded districts remains a challenge. The evidence so far comes from individual classroom experiments and policy announcements, not controlled studies on effectiveness phys.org. Some universities are also experimenting with AI-proof exam designs, such as embedding hidden prompts to catch cheating, but those approaches risk the same arms race Nature. The core insight for builders is that no single tool or policy will solve the problem. The most useful products will combine efficient triage (like Brisk) with assignment frameworks that naturally surface genuine understanding - and they'll need to be flexible enough that instructors can adapt them to their specific contexts.

The bottom line: AI has forced a fundamental rethink of how we measure learning. The institutions that figure this out first will define the next generation of educational assessment. For those building in the space, the key is to focus on process over output, even if that means building workflows that look less like traditional grading.