AI detectors make mistakes
AI detectors guess from the finished words alone. When Turnitin launched its detector, it said 1% of papers could be wrongly flagged. Vanderbilt University worked out that across the 75,000 papers its students submitted in a year, that could mean around 750 false flags, and turned the detector off.
A Stanford study found widely used detectors consistently misclassify writing by non-native English speakers as AI-generated.
What to do if you’re flagged
- Ask for the details. Which tool, what score, and which passages.
- Collect your evidence. Version history, drafts, notes and sources. See how to prove you didn’t use AI.
- Offer to explain the work. Walk through your argument and your sources in person.
- Read the policy. Check what your school or client says about AI and about detector results.
- Ask for a decision based on evidence. A detector score alone isn’t proof.
Next time, write it in Huma
Huma records how you type as you write. Press Finish and you get a link and a PDF Report showing a person typed the text, with how much was typed and how much was pasted. Your teacher, client or editor opens the link in any browser, with no account. Huma is free.
Questions
Can AI detectors be wrong?
Yes. They estimate from the words alone, and universities and researchers have reported false flags on human writing.
What is a Huma link?
A link to your text as you wrote it in Huma, with its Huma Score: how much was typed and how much was pasted. Anyone can open it in a browser and check it.
Will my teacher accept a Huma link?
That’s up to them, but they can check it themselves: the link shows the text and how it was written, with no account needed.