Detectors flag writing that is predictable, and some human writing is very predictable. Non-native English speakers, formal writers, and people writing in structured genres are flagged far more often than others.
What a false positive is
A false positive happens when a detector labels human writing as AI-generated. For the writer, it can mean a failing grade, a rejected submission, or a damaged reputation for work they did themselves.
The research on non-native English writers
In a 2023 study published in the journal Patterns, Stanford researchers ran seven popular detectors on 91 essays written by non-native English speakers for the TOEFL English test. The detectors labeled more than half of those human-written essays as AI-generated, and one detector flagged nearly 98% of them. The same detectors correctly classified more than 90% of essays by U.S. eighth-graders as human. The researchers traced the gap to perplexity: writers with a smaller English vocabulary use more common words, which detectors read as machine-like.
Who else gets flagged
- Writers who keep things simple and clear on purpose
- Students taught to follow strict structures, such as the five-paragraph essay
- Technical, legal, scientific, and business writers using standard phrasing
- People who use grammar and spelling tools heavily, which smooth out natural variation
- Neurodivergent writers whose style is consistent and precise
- Anyone writing a short piece, where there is little text to measure
Why it happens
Detectors do not detect AI. They detect predictability and uniformity, which AI text usually has. Plenty of human writing has those traits too. The detector has no way to tell a careful human writer from a language model when the finished text looks similar.
How to reduce the risk
- Keep your drafts, notes, and outlines.
- Write in tools that keep version history, such as Google Docs or Word with autosave, so you can show how your document developed.
- If you used a grammar tool or AI for brainstorming, know your school’s or client’s policy and be ready to explain exactly what you used it for.
- For work where authorship is likely to be questioned, write in a tool that produces verifiable evidence of your process.
If you’ve already been flagged
Stay calm, gather your evidence, and ask for the specific basis of the concern. Our guide for people falsely accused of using AI walks through each step.
Sources
Last reviewed October 2026 by the AblePro team. Found something out of date? Tell us.