A plagiarism finding says text matches a source without attribution. An AI flag says the text resembles generated text. One is a claim about provenance, the other a probability about texture, and they require completely different responses.
If a report shows similarity, the question is where the matching text came from and whether it is cited. If a report shows an AI indicator, the question is which sentences read uniform and whether the writer can show a process.
I have watched students try to answer an AI flag with a citation audit, and vice versa. Both times the real problem went unaddressed and the conversation got worse. Knowing which number you are arguing about is half the work.
This page separates the two Turnitin numbers, and this one explains acceptable AI percentages for the policy side.