Second-language writers
In 2023, Stanford researchers ran essays written by non-native English speakers for the TOEFL exam through seven popular AI detectors. The detectors falsely flagged an average of 61.3% of those essays as AI-generated — while flagging almost none of the essays by native speakers. The reason is built into how the tools work: they reward surprising word choices and punish predictable ones, and careful, correct, second-language English is — unfairly — predictable to a program.
If this is your case, it is not a detail. It is a central fact of your answer. State it plainly, cite the study, and pair it with your trail: drafts showing your sentences being built and revised, your notes, your earlier coursework in English. A panel that understands the tool's documented weakness reads your file differently.
The free checker has a question for exactly this — answer yes, and the Stanford evidence is placed in your ranked evidence list where it belongs, near the top.
Practical notes: if your school offers an international student office or an ombuds, tell them early; they have seen this pattern. If your writing process involved a dictionary, a grammar reference, or translation of single words, say so and show it — that is a normal writing process, documented in your drafts. And keep everything in your first language too: notes you took in it are evidence of a person thinking, not a program generating.
One honest limit: bias evidence strengthens an answer; it does not replace the trail. Build both — the study opens the panel's mind, and your drafts close the question.
Get the MyWorkMyWords Kit — $39
One payment of $39. No subscription. The kit is a web page you can return to and print — the evidence binder builder, appeal statement templates, hearing scripts, and deadline tracker, built for a student answering an accusation this week.