Turnitin’s AI Detection Is Losing Universities Fast — and Students Should Know Why

Published June 18, 2026 · Detection Drama

Turnitin’s AI Detection Is Losing Universities Fast — and Students Should Know Why

Illustration: Detection Drama · Universities worldwide are disabling Turnitin’s AI writing detection feature in 2026.

Curtin University confirmed in September 2025 that it would disable Turnitin’s AI writing detection feature across all campuses from January 1, 2026. Turnitin would stay on for plagiarism checking. The AI detection component — the part that flags student essays as “probably AI-written” — would be turned off. The university’s stated reason: “fostering trust and clarity within a modern academic culture.”

Curtin is not alone. A growing list of institutions including Vanderbilt University, the University of Cape Town, the University of Queensland, Johns Hopkins, and multiple University of California campuses have either deactivated or restricted Turnitin’s AI feature. In every case, the same concerns surface: false positives, equity issues — particularly for non-native English speakers — and a fundamental lack of transparency about how the algorithm works.

The numbers explain why institutions are walking away. Independent testing in 2026 finds GPTZero achieving roughly 84% accuracy with a 6–8% false positive rate; Turnitin lands at 85–90% accuracy but deliberately allows around 15% of AI content through to reduce false flags. These aren’t small margins. Applied across thousands of submitted essays, they mean wrongly accused students — regularly.

Dr Mark A. Bassett, Associate Professor and Academic Lead (Artificial Intelligence) at Charles Sturt University, was blunt on LinkedIn after Curtin’s announcement: “Kudos to Curtin University for joining the growing list of providers that are abandoning this deeply flawed technology.” That quote is from a senior AI academic at a teaching institution. It lands harder than a press release.

What makes this particularly awkward for Turnitin is that the company launched a new bypasser detection feature earlier — designed to catch AI text that had been run through humanizer tools — around the same time that universities were voting with their off-switches. Independent tests by academics including Northumbria University’s Tadhg Blommerde found significant variation in how different bypasser tools were flagged. Turnitin’s own response acknowledged that “no AI system can achieve zero false positives.” That’s a true statement. It’s also exactly the problem.

Where does this leave students? If you’re at an institution still running Turnitin’s AI detection, false positives are a documented risk — especially for writing that is highly structured, uses formal register, or was heavily edited. The threshold policies vary wildly: some professors treat any score above 20% as an investigation trigger; others ignore it entirely. You may not know which kind you have until it’s too late.

The counter-argument worth taking seriously: without any detection tool, academic integrity becomes much harder to enforce at scale. Redesigning assessments — oral defenses, in-class writing, personalized prompts — is the evidence-based alternative most researchers are advocating. But it requires more work from institutions than toggling a setting off. Not every department will do it.

The trajectory is clear. Turnitin’s AI detection feature launched with enormous fanfare and is exiting through the back door at institution after institution. For students currently facing allegations based on AI scores, that context matters. It should be part of any defense.