Before Your School Trusts Turnitin's AI Score: Six Questions to Ask (2026)

Verdict: A school that already runs Turnitin should treat the AI writing report as one signal in a conversation, not a finding. Before it relies on the report, it should ask the six questions below and get written answers from its vendor and its own legal office.

What the vendor itself says

FavTutor’s 2026 field review reports that Turnitin warns its AI Writing Report can misidentify human-written, AI-generated and AI-paraphrased text. The review also says Turnitin advises against using the report as the sole basis for any adverse action. That is a vendor position as the review describes it, and a school should read it in Turnitin’s own documentation before adopting a policy.

Turnitin’s homepage leads with institutional products and does not list a personal free AI check, as checked on 2026-10-14. FavTutor describes the product as generally unavailable as a standalone consumer detector, which matches that picture.

Source: FavTutor’s buyer notes.

Question 1: Will the score show a percentage?

For some results, the answer is no. FavTutor reports that Turnitin does not display a numeric percentage for results in the 1% to 19% range, because that band has a higher incidence of false positives. A school should ask how the report will look for a borderline essay, and what the staff member is expected to do with a result that has no number. Source: FavTutor’s review summary.

Question 2: How much text is needed before a report appears?

FavTutor reports that Turnitin requires a minimum amount of qualifying long-form prose before it produces an AI-writing report. A short reflection or a lab note may therefore receive no report at all. The school should confirm the minimum length in writing and decide what it will do with a short submission that gets none.

Question 3: What did the false-positive studies find?

eesel AI’s June 2026 review reports that a Stanford-linked study found Turnitin flagged 61.3% of essays by non-native English speakers as AI-generated. The review also states that Turnitin’s own under-1% false-positive claim applies only to documents where AI content exceeds 20% of the submission. A school should ask its vendor for the study, the population it covered, and the test conditions. Source: eesel AI’s spec summary.

Question 4: Which other schools have made different choices?

eesel AI’s review lists five universities that disabled the AI detection feature over false-positive concerns: UC Berkeley, Vanderbilt, Johns Hopkins, Michigan State and Northwestern. It also mentions UT Austin and Australian Catholic University as restricting or disabling it. Those decisions are the review’s report. A school should read the universities’ own policy pages before it cites them.

Question 5: What is the appeal process?

The review does not describe an appeal process, and this article does not invent one. A school should write its own procedure before a student is flagged. That procedure should say who reviews the report, what evidence the student can provide, such as drafts, notes or revision history, and how the outcome is recorded.

Question 6: Who is accountable for the decision?

Turnitin’s own warning, as FavTutor reports it, places the decision with the institution. The school should name a person who signs off on any action taken after an AI report. Without that step, the score can end up treated as a verdict, which the vendor says it is not.

What a reasonable policy looks like

• Treat the report as a prompt for a conversation with the student.
• Ask for drafts and revision history before any decision.
• Record the report, the date and the version used.
• Keep the report out of any decision that is not reviewed by a person.

What this article does not settle

No accuracy test of Turnitin was run for this article, so the rate of wrong reports is unknown here. It relies on FavTutor’s description of Turnitin’s warnings and on eesel AI’s description of the studies and university decisions. The school’s own policy and the vendor’s current documentation govern any real decision. For the method used on this site, see the method page.