Turnitin AI detection, explained
What the percentage means, what it doesn’t, and what to do if it lands on your essay. (VerifyWrite is independent and not affiliated with Turnitin.)
What Turnitin’s AI report shows
Since April 2023, Turnitin’s Similarity Report can include an AI writing indicator: the percentage of “qualifying text” – long-form prose sentences, not lists, tables or bibliographies – that its model believes was generated by a large language model. Highlighted segments show where.
- Scores of 20% and above are shown as a number with highlighted text.
- Scores of 1–19% are shown as an asterisk, because Turnitin found false positives more likely at low levels.
- A separate category flags text that looks AI-generated and then paraphrased by a word-spinner tool.
- Submissions need a minimum length (currently around 300 words of prose) for a result.
What it doesn’t show
It doesn’t show that a student cheated. Turnitin itself describes the indicator as a starting point for conversation and says it should not be used as the sole basis for adverse action. It doesn’t distinguish between permitted and forbidden AI use, and it doesn’t see your drafts.
How often is it wrong?
Turnitin has reported a document-level false-positive rate below 1% for documents over its minimum length with more than 20% AI writing detected, and a higher sentence-level rate. Independent testing and university reviews have questioned those figures, and several universities, including Vanderbilt, turned the feature off in 2023. Across thousands of submissions a year, even 1% means a lot of honest students – play with the base-rate simulator to see how many.
If your work is flagged
- Ask to see the report and which passages were highlighted.
- Gather your process evidence: version history, drafts, notes, sources you read.
- Offer to talk through the essay – your argument, why you chose your sources.
- Point out, politely, that Turnitin advises against relying on the score alone.
Checking before you submit
No free tool can tell you your Turnitin score. What you can do is look at your own writing the way a detector does: the VerifyWrite detector highlights sentences with AI-typical patterns and explains each one, so you can see if a passage reads as over-polished, and the plagiarism checker finds public-web matches before the similarity report does.
Questions people ask
How does Turnitin detect AI writing?
Turnitin splits a submission into overlapping segments of a few hundred words, scores each sentence with a classifier trained on human and AI text, and reports the share of qualifying prose it believes was AI-generated. It also has a separate indicator for text that appears to have been AI-generated and then run through a paraphrasing tool.
Can students see their Turnitin AI score?
By default the AI writing report is visible to instructors, not students. Institutions configure this, so check with your teacher or course handbook.
What does an asterisk (*%) mean in Turnitin’s AI report?
Turnitin shows scores between 1% and 19% as an asterisk because results in that range have a higher chance of false positives, so they are not displayed as a number.
Is a VerifyWrite score the same as a Turnitin score?
No. Different detectors use different methods and give different numbers for the same text. VerifyWrite shows transparent, explainable signals; it cannot predict what Turnitin will report.
Does Turnitin detect ChatGPT?
It is designed to detect text from ChatGPT and other large language models. Like every detector it misses some AI text, especially edited text, and it can wrongly flag human writing.
Check your work
Your text follows you between tools, so you can paste once and check everything.