Teachers: What AI Detection Score Do You Actually Act On?

Question for any educators on here. What Turnitin AI detection percentage actually triggers you to investigate further? I have heard wildly different answers.

My writing professor said she ignores anything under 40% because the false positive rate is too high at lower thresholds. My history professor apparently flags anything over 15%. And my CS professor doesnt even look at the AI scores at all because he says the tools cant handle technical writing.

From a student perspective this inconsistency is really stressful. I could submit the exact same writing quality to two classes and get flagged in one but not the other.

Would love to hear from teachers about what score you treat as a red flag and how you handle the conversation when you suspect AI use. Also do you use anything besides Turnitin? Some of my friends at other schools say their professors use GPTZero or Copyleaks instead.