Sometimes. Turnitin does not publish results for each AI model, and detectors trained mostly on ChatGPT-style text often miss newer models. In our own tests, Claude-written report sections were caught much less often by older detectors than ChatGPT text was.
Written by Adithyan, a student who builds PaperCheck. Last updated 22 October 2026.
We wrote 30 student-style report sections with Claude Haiku, Sonnet and Opus, kept them out of training, and scored them. Our detector flagged 27 of the 30 as mostly AI. An older detector we tested caught far fewer, and the hardest ones were literature reviews and methods sections.
Each AI model has its own habits. A detector learns the habits it was trained on, so when a new model writes differently, the score can drop to near zero even though the text is fully AI-written.
A low score on one tool does not mean another tool will agree. If you used Claude for parts of a report, read those parts again and make them about your own work.
Sometimes. Results vary by model and by section, and Turnitin does not publish per-model figures.
For many detectors, yes, because most were trained mainly on ChatGPT-style text.
Check your paper: 5 free checks a month
PaperCheck is not affiliated with Turnitin, GPTZero or Pangram. Scores from any detector are estimates, not proof.