The 88% Problem: Why Detection Fails and Verification Works

88% of UK students use GenAI. Turnitin flags 1 in 3 as suspicious. But 61% of those flags are false positives. It's time to admit detection isn't working.

Generative AI has fundamentally changed how students write, research, and synthesize information. In response, institutions deployed 'AI detectors'. However, emerging research shows these detectors disproportionately flag non-native English speakers. Instead of relying on flawed probability models, we must return to assessing authentic understanding through dialogue.

The verification approach is fundamentally different. Rather than asking "did a machine write this?", we ask "does this person understand what they submitted?" This shift — from detection to verification — is not merely semantic. It changes the entire relationship between student, institution, and assessment.