Cheat sheet for the video [Turnitin AI Detection: What a \*% Score Really Means](https://www.youtube.com/watch?v=b2slKVWFYzo) · [read the full page](https://kuzscotech.github.io/kush-exe/turnitin-ai-detection/)

## The short answer

Turnitin's AI detector estimates what percentage of a document's prose was likely written by AI, and highlights the sentences it predicts are AI-written. It is a prediction, not proof: Turnitin says its false positive rate is not zero and the score should not be the sole basis for action.

## What the detector actually measures

- The score is a percentage of qualifying prose sentences, not necessarily of the whole submission.
- A file needs at least 300 words of long-form prose to get an AI score. Turnitin says the detector is meant for prose, not poetry, scripts, code, bullet points or tables.
- From 1% to 19%, the report shows no number or highlights, only an asterisk (`*%`). Reports from before the July 2024 change may still show a number. Turnitin's guide says results below 20% are less reliable.

## Claims versus evidence

| Claim | What the evidence says |
|---|---|
| `*%` means no AI was found | No. Something was flagged, between 1% and 19%. 0% is a separate result. |
| `*%` proves a little AI use | Not proof. Turnitin hides the number to avoid false positives. |
| False positives are under 1% | Turnitin states that only for documents with over 20% AI writing: about 1 in 100 fully human-written documents. |
| Every highlighted sentence is AI | By Turnitin's 2023 figures, about 4 in 100 highlighted sentences may be human-written. |

## What it cannot tell

Turnitin's 2023 figures show where wrong flags cluster. The model has been updated since.

- **Next to AI text.** 54% of falsely flagged sentences sat right beside actual AI writing. Turnitin says an October 2026 update for English submissions targets these transition points; older reports are not rescored unless resubmitted.
- **Introductions and conclusions.** More false positives appeared in the first and last few sentences.
- **Certain human writing.** Turnitin's current FAQ says false positives can include text with little structural variation, repetitive text, and paraphrase that develops no new ideas.

Vanderbilt disabled the detector in August 2023, estimating that a 1% false positive rate could have wrongly labeled about 750 of its 75,000 papers from 2022. Curtin said it would disable it from 1 January 2026.

## If your work is flagged

Turnitin's guidance says:

- Turnitin makes no determination of misconduct; the final decision rests with the instructor.
- The score is a single data point, not a definitive answer in isolation.
- Students do not see the AI report by default. Instructors can share it as a PDF or make it available.

Sources: [Turnitin help center](https://helpcenter.turnitin.com/hc/en-us/articles/46245332050317-Why-is-the-AI-Writing-Detection-report-score-showing-as), [Turnitin FAQ](https://guides.turnitin.com/hc/en-us/articles/28477544839821-Turnitin-s-AI-writing-detection-capabilities-FAQs), [Turnitin guide](https://guides.turnitin.com/hc/en-us/articles/22774058814093-Using-the-AI-Writing-Report), [Turnitin update, May 2023](https://www.turnitin.ca/blog/ai-writing-detection-update-from-turnitins-chief-product-officer), [Vanderbilt](https://www.vanderbilt.edu/brightspace/2023/08/16/guidance-on-ai-detection-and-why-were-disabling-turnitins-ai-detector/), [Curtin](https://www.curtin.edu.au/news/oasis-news/update-on-turnitin-ai-detection-tool/).