Talking with a writer about a flagged result
A fair, five-step way for teachers and editors to follow up on a low human score without accusing anyone.
Updated Oct 8, 20261 min read
A detector can tell you where to look. It can't tell you what happened. The fairest follow-up treats the scan as one piece of evidence and gives the writer a real chance to explain.
- Step 1: Re-read the work yourself. Note what stood out before you saw the score: shifts in voice, generic claims, sources that don't check out.
- Step 2: Check where the highlights are. Are they clustered in one section, or scattered single sentences? Clusters mean more.
- Step 3: Compare with earlier work. A sudden change in style, vocabulary or quality tells you more than any single score.
- Step 4: Ask open questions. “Walk me through how you wrote this section” works far better than “Did you use ChatGPT?”
- Step 5: Ask for process evidence. Drafts, version history, notes and outlines usually settle the question quickly.
Things to avoid
- Presenting a score as proof in a disciplinary process without other evidence.
- Re-scanning until you get the result you expected. The same text gives the same score; changing the text changes the result.
- Comparing scores between different detectors. They use different models and different scales.