Your Turnitin AI score, explained
It is the estimated share of scoreable sentences that Turnitin’s model rated as likely machine-written. It is not a certainty, not a plagiarism figure, and not a record of what you did.
It is not the similarity score
Two separate systems, two separate numbers. Similarity compares your text against existing documents. The AI indicator compares it against learned patterns of machine prose. A paper can show 0% similarity and 100% AI, or the reverse, neither constrains the other.
What counts as high
There is no official threshold. Any specific cutoff you have been told is your institution’s policy, not a property of the tool. In practice most institutions start asking questions somewhere around 20% and treat scores above roughly 50% as needing an explanation.
Where it is least reliable
- Documents under a few hundred words
- Writing by non-native English speakers: a documented over-flagging group
- Heavily edited human text, where revision removed the natural unevenness
- Lists, tables and fragments, which are largely unscoreable
Check how smooth your writing reads
This measures the same surface properties detectors weight: sentence variation, vocabulary range, assistant phrasing, and shows you which ones stand out. It will not tell you what Turnitin will say.