Turnitin says yes: its AI writing detection is built to flag text that AI tools such as ChatGPT may have generated or modified. It reports a percentage and sentence highlights for instructors, and states its own limits. Everything below is what Turnitin itself says, from its guides, checked on 10 November 2026.
What Turnitin says it detects
- An overall percentage of qualifying text that AI writing tools or large language models may have generated or modified, with highlighted sentences. Each sentence gets a score, and the document score is the share of sentences over a threshold.
- Text rewritten by AI paraphrasers and bypass tools, in English only.
- Long-form prose only. It skips lists, headers, references, tables and code, and says it does not reliably detect non-prose or code.
- It does not name which model wrote a passage.
Sources: Turnitin AI writing detection FAQ, updated 3 November 2026, and its AI writing detection release notes.
Stated limits
- At least 300 words of prose are required, and Turnitin says shorter documents give a likely less accurate score. Files can be up to 30,000 words.
- Languages: English, Spanish, Japanese, Modern Arabic, Dutch, German and Turkish. Other languages are not processed.
- Turnitin says it aims to keep false positives on fully human text under 1% for documents with over 20% AI writing, and says it can miss some AI text.
- Scores of 1% to 19% are not shown, because Turnitin says those results are less reliable and false positives are more likely.
- It lists likely false-positive content: text without much structural variation, text that repeats itself, and text paraphrased without new ideas.
- Turnitin says the score should not be used as the sole basis for action.
Who can use it
Turnitin's AI writing report sits inside its institutional products. There is no public paste box and no URL input, and individuals have no free tier. Instructors and administrators see the indicator. The ChatGPT watermark from OpenAI is a different matter: it covers EU text only and its detector is restricted.
What an independent second check adds
A second detector built on different data gives you a cross-check before a conversation with a student or an author.
- It works on pasted text from the first 100 words, without an institutional licence.
- It shows writing-style similarity to ChatGPT and other families and marks the passages, so you can read the evidence yourself.
- It reads French, German, Spanish, Chinese, Vietnamese and Arabic as well as English.
ChatGPTDetect runs on Fileverity's Veread engine. In tests on human-written text across seven languages it wrongly flagged 0.5%; the facts page shows how that was measured. For classroom practice, read AI detection for teachers.