The other detectors

Flagged by another detector? The questions are the same.

Copyleaks looks for statistical patterns in phrasing and claims accuracy above 99% with a false-positive rate of 0.2% — its own figures. Its own best-practices guide still tells educators to “use the AI report as one data point” in a conversation about how the student worked. Bloomberg's pre-ChatGPT essay test included Copyleaks: it falsely flagged human essays at 1–2%, some at near-total certainty.

Originality.ai returns a 0–100 score with sentence breakdowns and advertises model-by-model accuracy around 99% with false-positive rates between 0.5% and 1.5% — again, the vendor grading its own homework. ZeroGPT gives simpler classifications with self-reported accuracy and no peer review behind it that we could find. None of these numbers was produced by your school, on your paper, in your discipline.

The independent picture: the most-cited academic test of the category (Weber-Wulff and colleagues, 2023) ran 14 detection tools and found none above 80% accuracy, concluding the tools were neither accurate nor reliable for exactly the use your school is making of one. That study is dated — the tools have changed since — but no vendor has published independent evidence that overturns its caution, and every vendor's own documentation repeats it.

Whatever the tool, ask the same five questions, in writing: Which program and which version? What is the exact score and the full report? What false-positive rate does the vendor publish, and on what test? Does our school's own policy limit scores as sole evidence (check the table)? And who decides, under what standard? A case that cannot answer its own questions is not a case — it is a number looking for a person to agree with it.

Then build the thing that does answer: your trail, from the evidence list, and your deadline from the checker (choose “another detector”).

Sources: Copyleaks pattern-recognition and best-practices pages; Originality.ai 2025 review and false-positives pages; Weber-Wulff et al., International Journal for Educational Integrity, 2023 — all checked October 19, 2026. Vendor accuracy figures are the vendors' own claims and are labeled as such wherever this site repeats them.