Turnitin

Turnitin flagged your essay. Here is what the number is.

The percentage is a share of sentences, not a verdict. In Turnitin's own description, its model scores each sentence, labels sentences above a threshold as AI, and the document score is the percentage of those qualifying sentences flagged. It counts only prose sentences — lists, references, tables, and formulas are excluded — and papers with under 300 words of prose aren't scored at all. A score of 87% means 87% of the counted sentences crossed a pattern threshold. It does not mean anyone proved who typed them.

The asterisk range is the tell. For scores from 1% to 19%, Turnitin shows no percentage and no highlights at all — just an asterisk — in its own words, “to avoid potential incidence of false positives.” If your accusation involves a low score, or a score nobody will show you the basis for, ask how a range the vendor itself suppresses became evidence.

The error rates, in Turnitin's own numbers: the headline claim is a false-positive rate under 1% — but defined only for documents with over 20% of text flagged, which quietly excludes the borderland where accusations live. At sentence level, Turnitin states its false-positive rate is around 4%: roughly 1 in every 25 highlighted sentences is expected to be a mistake. Turnitin validated its headline rate on over 700,000 pre-ChatGPT papers — and even at 1%, a big university's year of papers yields hundreds of wrongly flagged students, which is why Vanderbilt did the math and switched the tool off.

Turnitin's own sentence, worth quoting in your answer: “the percentage on the AI writing indicator should not be used as the sole basis for action or a definitive grading measure by instructors,” and “Turnitin does not make a determination of misconduct.” The vendor sells a conversation-starter. If your school is treating it as a conclusion, your statement can say so — with the vendor as your source.

What to ask for: the complete AI writing report, not the number — which sentences were flagged, in which sections, and whether your score was ever in the asterisk range. Then answer with the trail: the evidence ranking exists because a percentage cannot cross-examine a version history.

Run your dates through the checker — choose Turnitin and enter your score, including if it's in the asterisk range, and the checker will place the right vendor facts into your case summary.

Sources: Turnitin's own AI writing detection FAQ (guides.turnitin.com), checked October 19, 2026; Turnitin's published false-positive explanation (turnitin.ca).