New: Institutional Licensing, deploy across your district or college. Read the framework →
A aiessaydetector.ai

Pillar guide · for ESL writers

Your English is fine. The detector is broken.

Why non-native English writers keep getting flagged as AI, what the research actually says, and a practical list of things to do before you submit.

Published 2026-03-10 · Updated 2026-04-05 · Editorial Team

The numbers.

Liang et al. (2023), Patterns: essays by non-native English speakers were misclassified as AI-generated by commercial detectors at rates up to 61%, versus under 5% for native speakers. Replicated multiple times since.

Our own published number (on /stats): 7.2% ESL false-positive rate vs. 3.1% native-speaker false-positive rate. Better than most commercial detectors, but still a 2.3x gap.

The gap isn't random. It has a specific cause: the formal academic register ESL writers are most frequently taught in school produces text with low burstiness and low perplexity, the same statistical signature that fine-tuned LLMs produce.

Why it isn't your fault.

You were taught to write formally. You were taught to avoid contractions, to vary word choice, to use structured transitions, to cite carefully, to use passive voice for objectivity. Every one of those teaching directives pushes your prose in the direction that AI detectors flag. This is a documented failure of the detectors, not a reflection of your writing quality.

Your professor may not know this. Your institution's academic-integrity policy may not acknowledge it. That's changing, as of 2026, more institutions explicitly name ESL false-positive bias in their policies, but coverage is uneven.

What to do before you submit.

  1. Save your draft history. Always. Google Docs version history is the strongest single piece of evidence you can produce if flagged. Never write a final essay outside of a tool with version history.
  2. Write a short disclosure paragraph if your institution allows AI-assisted editing. "I used a grammar-checker / translation tool for [sections], here's what it helped with." This removes the ambiguity that detector flags feed on.
  3. Add one or two personal observations. An "I remember…" sentence or a reflection on your own experience is hard to fake and easy to defend in a conversation.
  4. Vary sentence length deliberately. A 5-word sentence followed by a 25-word one is the single fastest way to reduce detector signal without changing your voice.
  5. Keep receipts on your sources. If you can produce notes from the books and articles you cited, it's hard to argue you didn't write the essay yourself.

If you're flagged anyway.

Don't panic. Don't confess to something you didn't do. The conversation that most reliably works, based on the student reports we've heard:

  1. Ask for the specific passages that were flagged, not just the overall percentage.
  2. Walk your professor through your draft history.
  3. Cite Liang et al. (2023) if they don't already know about the ESL false-positive problem, it's a well-known paper in the field.
  4. If the conversation doesn't resolve, escalate politely to the department chair or academic-integrity office.

It's worth saying: the system isn't fair to you right now. You're going to have to advocate for yourself in ways native-English students don't have to. That's a real cost, and it isn't yours to absorb silently.

Frequently asked questions

Can I just use a humanizer to avoid false flags?
Ethically, yes, for human-written text being misclassified. Our humanizer's whole design purpose includes this case. That said, always save your draft history and always disclose AI-assisted editing if your institution requires it.
Are there detectors that don't have this bias?
Every commercial detector has some version of this bias. Some are better than others. We publish our numbers; most don't. A detector that doesn't publish ESL false-positive numbers almost certainly has worse ones than we do.

Check an essay now.

Open the detector →