HomeCase studies › Case 01
Case 01 · Equity crisis

Turnitin flags non-native English writers as AI at 61%.

The failure that got Turnitin removed from a dozen universities.

The failure · on the record

Turnitin's AI detector flags non-native English writers as AI-generated at up to 61% — nearly 3× the rate for native speakers. In May 2026 a California family escalated a 76% Turnitin flag into federal court. A New York court reversed a student expulsion over a false Turnitin AI flag. American, Boston U, UC Berkeley, Colorado State, DePaul, Georgetown, Michigan State, NYU, University of Cape Town, and Curtin (as of Jan 1 2026) have quietly disabled Turnitin AI detection.

The numbers
PopulationFalse-positive rateSource
Native English writers5–20%casrai.org 2026
Non-native (ESL) writersup to 61%casrai.org 2026 / Turnitin internal
Truth-in-AI on German paraphrase0.4% (verified)RAID.signed.json
What we do differently

Signed proof of 99.6% AUC on German-language paraphrase attack — the same adversarial condition that breaks every commercial detector on non-English text.

Bring your own ESL corpus. We return an Ed25519‑signed score per document.

Proof
See the German paraphrase result verified in-browser. Open the verifier →
Try it yourself
Try a live scoring call.
curl -X POST https://api.trigeochiral.com/detect \
  -H 'Authorization: Bearer $TRIALKEY' \
  -d '{"text": "Paste any ESL essay here"}'
Bounty · put money where the claim is
$1,000
Per legitimate ESL essay we false-positive as AI at score ≥ 0.85. Cap 10 payouts per organization. Submit at /challenge/.
Sources