Teacher’s guide to actually understanding AI detection reports

I have seen too many colleagues misinterpret AI detection reports so I wrote this guide. Share it with any teacher who needs it.

What the overall percentage means:
It is NOT a confidence score. An 80% does not mean 80% chance of AI. It means the tool thinks roughly 80% of the text was AI-generated. A student could write a great intro and conclusion (20%) and use AI for the body (80%).

Highlighted text matters more than the score:
Look at WHICH sections are highlighted. If the methodology section is flagged but the literature review is not, the student probably used AI to help structure their argument but did the research themselves. Context matters.

Score ranges and what to do:
– 0-15%: Likely human-written. Normal variation in writing can trigger this.
– 15-30%: Gray zone. Could be light AI assistance or just a structured writing style. Monitor but do not accuse.
– 30-60%: Warrants a conversation with the student. Ask about their process.
– 60%+: Strong indicator of AI use. Have a formal conversation.

NEVER DO THIS:
– Accuse a student based solely on a detection score
– Share detection scores publicly
– Use detection as the only evidence in an integrity hearing
– Assume ESL students who get flagged are cheating

The detection score is a starting point for conversation, not a verdict.