Group projects and AI detection — whose score is it?
A detector scores a document, not a person. Group work makes that gap matter.
Scan sections, not just the whole document
If your platform allows it, scan each student’s contributed section separately rather than the merged whole — version history in Google Docs can usually attribute paragraphs to an author even after merging. A single high score on a combined document tells you nothing about which student’s writing produced it.
Don’t assume the score applies evenly
One student is at least as likely to have written the flagged section as any other, and a score on 3,000 combined words says little about any single 500-word contribution. Read section by section before you form a view.
Ask the group together, then individually
A group conversation surfaces who wrote what. If one student’s account doesn’t match what the version history shows, a private follow-up with just that student is fairer to the rest of the group than raising it in front of everyone.
The policy question worth settling upfront
Decide, and tell students before the project starts, whether AI-assistance rules apply per section or to the group as a whole — see setting a policy students will actually read. Groups negotiate this poorly themselves if the rule was never stated.