Does Turnitin detect Gemini?
Yes
Yes. The detector scores the statistical shape of the writing rather than matching against known outputs, and Turnitin's own model documentation lists Gemini versions among what its English detector covers.
How it works
Each sentence is scored for how predictable the word choices are and how little the rhythm varies. Every instruction-tuned assistant produces that profile, so no list of models is needed for the number to move. Gemini has one wrinkle the others do not: it is wired into Google Docs, so people use it inside the document they are writing rather than pasting from a chat window. The AI report cannot see that, but the revision history can, because generated text arrives as one large insertion.
The part people miss
The report never names a model, so it cannot separate Gemini from ChatGPT from your own even prose. Turnitin also needs at least 300 words of continuous prose before it produces an AI report at all. Google says Gemini output carries a SynthID watermark, but that is readable by Google rather than by Turnitin, and there is no public way for a student or an instructor to check a passage of text for it.
Flagged for something you wrote yourself?
There is a right order to handle it in, and editing the document first destroys your best evidence.