Universities Abandon AI Detectors: New Methods Emerge for Evaluating Students

A growing number of universities in the United States have decided to ban the use of artificial intelligence detectors in academic assessment. The main reason cited is that these tools produce "unreliable" results and can lead to unfounded accusations of cheating, according to Inside Higher Ed.

A controversial tool

AI detectors were quickly adopted by higher education institutions following the explosive rise of ChatGPT and other generative models in student work. However, practical experience has shown that these systems frequently generate "false positives," flagging text written entirely by students as AI-generated—particularly when those students are not native English speakers or have a more rigid writing style.

This lack of accuracy has led several universities to ban the use of such detectors as sole evidence in cases of suspected academic fraud, in order to avoid wrongly penalizing students.

A dilemma for teachers

Abandoning detectors, however, leaves faculty in a difficult position. Many instructors say the improper use of artificial intelligence in academic work is widespread, but they no longer have a technical tool to prove it.

According to Inside Higher Ed, this situation is forcing institutions to fundamentally rethink how they assess students, shifting the focus from detecting fraud to preventing it through pedagogical design.

New forms of assessment

Against this backdrop, universities are exploring alternative methods of testing knowledge: oral exams, real-time presentations, supervised in-class writing assignments, and projects that require explicit documentation of the working process, not just the final result.

These approaches aim to reduce reliance on automated detection and instead emphasize students' genuine ability to demonstrate understanding and critical thinking, regardless of the digital tools available to them.

A paradigm shift

The American case reflects a broader trend in higher education: adapting assessment methods to a reality in which artificial intelligence is omnipresent. Rather than searching for a perfect technological solution to detect AI use, many institutions appear to be opting instead for a complete redesign of how learning is tested.