AI detector false positives
Getting told your own work "looks AI-generated" is a rotten feeling. It's also more common than most people think. Here's why it happens, who it happens to most, and what to do about it.
Updated October 10, 2026 ยท By the IsThisAI? team
What a false positive is
A false positive is when a detector labels human writing as AI-written. Every detector produces them. The question is how often, and on what kind of writing.
Turnitin, the most widely used academic checker, aims to keep its document-level false positive rate under 1% for documents with over 20% AI writing. Put plainly, it "might flag a human-written document as AI-written for one out of every 100 fully-human written documents" (Turnitin FAQ). At the sentence level, Turnitin has said the rate is around 4%, and that it's more common where human and AI text are mixed (Turnitin blog).
Other tools have done worse. OpenAI's own classifier wrongly labeled human writing as AI 9% of the time before OpenAI withdrew it in 2023 (OpenAI).
Want a quick second opinion on a piece of writing? Paste it into IsThisAI? and see which sentences read as AI-like.
Check your text free →Who gets flagged most
Non-native English writers
The clearest evidence comes from a 2023 Stanford study. Researchers ran seven popular detectors on 91 TOEFL essays written by people and 88 essays by US eighth graders. The detectors were near-perfect on the eighth-grade essays but labeled more than half of the TOEFL essays AI-generated, with an average false positive rate of 61.3%. All seven tools agreed on 19.8% of the TOEFL essays being AI, and at least one tool flagged 97.8% of them (Liang et al., 2023).
The reason is word choice. Writers with a smaller English vocabulary tend to use common, predictable words, and predictability is exactly what many detectors read as "AI." When the researchers had ChatGPT make the TOEFL essays' vocabulary fancier, the false positive rate dropped to 11.6% (Liang et al., 2023). Vanderbilt cited this bias when it disabled Turnitin's AI detection (Vanderbilt).
Formal, formulaic, or repetitive writing
Turnitin lists the kinds of text that show up in its own false positives: "content without a lot of structural variation, text that literally repeats itself, or text that has been paraphrased without developing new ideas" (Turnitin FAQ). Lab reports, legal-style writing, five-paragraph essays built from a template, and summaries of a reading can all fit that description.
Short pieces
On short documents, Turnitin says its prediction is "mostly all or nothing," so a mix of original and AI text "could be flagged as entirely AI-generated" (Turnitin FAQ). Short text gives any detector less to go on.
Turnitin false positives specifically
A few Turnitin details are worth knowing if you've been flagged:
- Low scores are hidden. Scores from 1% to 19% show only as an asterisk with no highlights, because false positives are more likely in that range (Turnitin FAQ).
- Only prose counts. Bullet points, lists, and code are not analyzed, so the percentage is of the "qualifying" prose, not the whole file (Turnitin FAQ).
- Turnitin says not to use it alone. Its guide says the model "may not always be accurate" and "should not be used as the sole basis for adverse actions against a student" (Turnitin Guides).
- Students usually can't see the report. Only instructors and admins see the AI indicator, though instructors can download and share it as a PDF (Turnitin FAQ). It's fair to ask for a copy.
Some schools have decided the tradeoff isn't worth it. Vanderbilt turned Turnitin's AI detector off in August 2023, citing false positives, bias against non-native writers, and a lack of transparency about how it works (Vanderbilt).
What to do if it happens to you
- Stay calm and get the details. Ask which tool was used, what the score was, and which passages were highlighted. Ask for the report PDF.
- Gather your process evidence. Version history in Google Docs or Word, earlier drafts, outlines, notes, research tabs, and emails to classmates or tutors all show how the work came together. Our guide on how to prove you didn't use AI walks through each one.
- Offer to talk it through. Being able to explain your argument, your sources, and why you made certain choices is strong evidence that the work is yours.
- Point to the published limits. Turnitin's own guidance, the Liang study, and the Weber-Wulff review (which found the detectors they tested unreliable, especially on edited text) are all fair to cite politely (Weber-Wulff et al.).
- Know your school's process. Find the academic integrity policy and follow its appeal steps. Many schools have an ombuds office or student advocate who can help.
Check any essay, email, or post for AI-sounding writing. Free, no sign-up, and your text stays on your device.
Check your text free →FAQ
How common are AI detector false positives?
It depends on the tool and the writing. Turnitin targets under 1% at the document level (for documents with over 20% AI writing) and has reported about 4% at the sentence level. Independent research found far higher rates for some groups, such as an average 61.3% on essays by non-native English writers across seven detectors.
Can Turnitin be wrong about AI?
Yes. Turnitin itself says its AI writing model may misidentify human-written text and should not be the sole basis for action against a student.
Why does formal writing get flagged as AI?
Detectors look for predictable word choices and low variation. Formal, templated, or repetitive writing has those traits, so it can look AI-like even when a person wrote it.
Should I run my essay through a detector before submitting?
It can tell you if your writing reads as robotic, but a low score elsewhere doesn't guarantee a low score in your school's tool. The best protection is keeping drafts and version history.
More guides
Sources
- Turnitin, "Turnitin's AI writing detection capabilities FAQs" (Turnitin Guides, updated Oct 7, 2026). https://www.turnitin.com/products/features/ai-writing-detection/faq
- Turnitin, "Understanding AI writing detection: False positive rates" (Turnitin blog, 2023). https://www.turnitin.com/blog/understanding-the-false-positive-rate-for-sentences-of-our-ai-writing-detection-capability
- Turnitin, "Using the AI Writing Report" (Turnitin Guides). https://guides.turnitin.com/hc/en-us/articles/22774058814093-Using-the-AI-Writing-Report
- Liang, Yuksekgonul, Mao, Wu & Zou, "GPT detectors are biased against non-native English writers," Patterns 4(7), 2023. https://www.cell.com/patterns/fulltext/S2666-3899(23)00130-7
- Vanderbilt University Brightspace team, "Guidance on AI Detection and Why We're Disabling Turnitin's AI Detector" (Aug 16, 2023). https://www.vanderbilt.edu/brightspace/2023/08/16/guidance-on-ai-detection-and-why-were-disabling-turnitins-ai-detector/
- OpenAI, "New AI classifier for indicating AI-written text" (Jan 2023; updated July 20, 2023 when the tool was withdrawn). https://openai.com/index/new-ai-classifier-for-indicating-ai-written-text/
- Weber-Wulff et al., "Testing of detection tools for AI-generated text," International Journal for Educational Integrity 19, 26 (2023). https://link.springer.com/article/10.1007/s40979-023-00146-z
This guide is general information, not legal or academic-policy advice. AI detection results, including ours, can be wrong.