The short answer
Nobody outside QuillBot has shown a reliable overall figure. QuillBot cites a 99% detection rate. One small preprint found it caught all 30 AI essays but flagged 10 of 30 human ones. Another found near-perfect results on unedited AI text, then much weaker results once the text was rewritten. Both are small, so treat them as warning signs and not as measurements.
What QuillBot claims
Its AI detector page cites a 99% detection rate from independent RAID evaluations and says its models are built to minimize false positives. RAID is a third-party benchmark. Other makers cite it for their own tools too, and we have not checked any of these results. RAID’s texts are not your essay, so the figure may not carry over.
The same page and its help centre also state limits:
- Length. At least 80 words. Texts of 300 words or more give more reliable scores.
- Formulaic human writing. Academic definitions, legal copy and structured templates can score higher than expected.
- Edited text. Heavily paraphrased or lightly AI-edited text is harder to classify.
- Not a verdict. It says never to rely on a detector alone for decisions that could affect someone’s career or academic standing. Its help centre says no tool is perfect.
What published tests found
| Study | Texts | QuillBot result |
|---|---|---|
| Yeadon et al., 2024 (preprint) | 30 AI and 30 human short physics essays | 83% accuracy. All 30 AI essays found. 10 human essays flagged as AI. |
| Alshammari and Rao, 2025 (preprint) | 49 human and 49 DeepSeek samples, plus 196 rewritten versions | Near-perfect on unedited and paraphrased AI text. 58% accuracy on text put through a rewriting tool. |
The first paper is from 2024, and QuillBot may have changed its model since. Neither paper is peer reviewed, as far as we could see. We found no large independent test of QuillBot’s detector. If you find a figure on a review site, look for the study behind it.
Why a QuillBot score is not proof
A score here is one company’s estimate. QuillBot’s own page says so. It is a reason to look closer, and nothing more.
Three things hold for any AI detector, this one included:
- It reads the finished text. It does not see how the text was written, so it can only estimate.
- Its figures depend on the test. An accuracy number holds for the texts it was measured on. Your text may be shorter, edited or in another language.
- A small error rate is still real people. Liang and colleagues (2023) found seven detectors, which the paper does not name, flagged 61.3% of TOEFL essays by non-native speakers on average. Detectors have changed since, but plain, predictable writing remains the risk.
What to do with a result
- If you are a student. Read your course policy, and keep your drafts and notes. See how to prove you didn’t use AI.
- If you are a teacher. Look at drafts and talk to the student before you decide. See how teachers check for AI.
Our AI detector gives a separate estimate for every sentence. It can disagree with QuillBot. We publish no accuracy figure for it.
Sources
- QuillBot: AI detector page and help article on false positives (page text seen in search results, 1 November 2026), and free or premium (read 10 October 2026).
- Studies: Yeadon et al. and Alshammari and Rao.