How often is Turnitin's AI detector wrong?
What Turnitin says about its own accuracy, what independent research found, which students get flagged most, and why universities are changing course in 2026.
If you are facing an AI accusation, the question is not whether the detector is usually right. It is whether it can be wrong about you. The answer, from Turnitin and from independent research, is yes.
What Turnitin says about itself
Turnitin's own guidance for teachers includes these points, in its words:
- The detector "may misidentify human-written, AI-generated, and AI-paraphrased text".
- It "should not be used as the sole basis for adverse actions against a student".
- "There is a higher incidence of false positives when the percentage is between 0 and 19", which is why scores in that range show as a star instead of a number.
- The percentage covers only "qualifying text", long-form prose, and "is not necessarily the percentage of the entire submission".
Turnitin has also claimed a document-level false positive rate below 1% at the 20% line. Both things can be true: a low rate across millions of papers still means thousands of wrongly flagged students every term.
What independent research found
- A Stanford study tested seven widely used detectors on essays written for the TOEFL exam by non-native English speakers. The detectors wrongly flagged 61% of that human writing as AI. On essays by US students writing in their first language, the same tools were far more accurate. The Higher Education Policy Institute summarised the evidence in August 2026 and called it "the fairness gap".
- Studies of EFL student writing concluded that neither Turnitin nor other commercial detectors "can currently be considered sufficiently reliable as AI detection solutions in academic settings".
- Detectors are also easy to fool in the other direction. The Stanford team found the fastest way to pass was to have an AI rewrite the essay. So the tools miss cheats and catch honest writers; that is the worst combination for a student relying on them.
Who gets flagged most
Formal writers, technical writers, non-native speakers, students who use grammar tools that rewrite sentences, and anyone whose style is regular and tidy. Several students have described being told their work was flagged because it was "too advanced" or "too polished".
What universities are doing in 2026
- In Australia, major universities have moved away from AI detection software after wrongful accusations, and an academic integrity director at Charles Sturt University has argued the sector should stop using it altogether.
- In the UK, universities forced Turnitin to pause a new licence that would have widened how student work is processed, and a St Andrews medical student's case made national news after he was accused over a dissertation he wrote himself.
- Students facing accusations are hiring lawyers; the Financial Times reported in October 2026 that one spent over £3,000 clearing their name with drafts and notes.
None of this means your school will drop the score. Most US and UK schools still use it. It means the score is contested, and a calm, documented response usually wins.
Sources
- Turnitin, "Using the AI Writing Report" (guides.turnitin.com)
- HEPI, "AI detectors and the fairness gap", 27 August 2026
- BBC News, "I was accused of using AI in my dissertation but I wrote it all myself", 10 August 2026
- news.com.au, "Major universities abandon AI detection tools amid fears of false accusations", 13 August 2026
- Charles Sturt University, "Unis and schools are moving away from AI detection software", 24 September 2026
Need the report of your own draft?
A Turnitin check shows the AI score and the flagged sentences before anyone else sees them. $4, both reports, the submission ID on every page.