Can Turnitin detect ChatGPT? What students should actually expect
Turnitin sells an AI writing detection feature to institutions, and it reports an estimated percentage of text it associates with AI generation. It is not a plagiarism match and it is not proof of how you wrote. Turnitin instructs institutions to treat the figure as a starting point for human review. Students cannot run Turnitin themselves, and no third-party tool can tell you what an institution’s Turnitin instance will report.
What does Turnitin’s AI indicator actually report?
Turnitin added an AI writing indicator alongside its long-standing similarity report. The two are different measurements. Similarity compares your text against a database of existing sources. The AI indicator is a prediction about writing patterns, with no matching source behind it.
The indicator appears as an estimated percentage of the document the model associates with AI generation, and in most configurations it is visible to instructors and administrators rather than to students. Turnitin’s own guidance describes the score as something that requires interpretation by a person who knows the assignment and the student.
Drawn from [1]
Can I run Turnitin on my own essay first?
Usually not. Turnitin is sold to institutions rather than individuals, and access runs through your school’s learning management system. Some universities enable a draft-check area; many do not. Your instructor or library is the only reliable source of truth about whether a pre-check exists in your course.
Be skeptical of any website claiming to give you “your real Turnitin score.” Those services are not running your institution’s Turnitin configuration, and submitting a draft to an unknown third party can create its own integrity problem if the text later appears in another database.
Does Cortex run Turnitin?
No. Cortex does not run Turnitin, does not partner with Turnitin and cannot predict a Turnitin result. Cortex’s detector connects to GPTZero when configured, which is a different company using a different model and reporting a different quantity.
Treating a GPTZero estimate as a Turnitin preview is a category error. Two detectors can disagree about the same paragraph because they were trained differently and calibrated to different thresholds. Use any estimate as a prompt to reread your own work, not as a forecast of an institutional result.
Drawn from [2]
How accurate is it?
Turnitin has published its own accuracy claims, and independent researchers have published results that differ. Both can be partly true, because accuracy depends on the test set, document length, the mix of human and AI writing, and where the decision threshold sits.
Two failure modes matter to you specifically. False positives flag writing you actually wrote. False negatives miss generated text. A vendor can reduce one at the cost of the other, so a single headline percentage tells you very little without the methodology behind it.
- Ask which product version and date an accuracy claim refers to.
- Check whether the test included edited and mixed-authorship writing.
- Look for false-positive rates, not only overall accuracy.
- Treat vendor-run and competitor-run tests with equal caution.
What do I do if Turnitin flags my essay?
Do not start rewriting to chase a lower number. Rewriting destroys the evidence that actually helps you, which is your genuine drafting history. Keep the original file, your notes, your sources and any version history your editor saved.
Ask your instructor which passage raised the concern and what the review process involves. Explain honestly how you wrote the piece, including any assistance you used and whether it was permitted. A flag opens a conversation; it does not settle one.
- Save the unedited draft and any version history.
- Collect research notes, outlines and source annotations.
- Read the syllabus policy on AI assistance before replying.
- Ask which specific passages were flagged.
- Describe the help you used, accurately.
Drawn from [3]
Common questions
Can Turnitin tell if I used ChatGPT?
It produces an estimate, not proof. Turnitin’s AI indicator predicts whether text resembles AI-generated writing, and Turnitin instructs institutions to treat it as a starting point for human review rather than a verdict.
Will rewording my essay remove a Turnitin AI flag?
There is no reliable way to know, and chasing the number is the wrong goal. If the assistance you used was not permitted, rewording does not resolve the policy question. If the work is genuinely yours, your drafts and notes are stronger evidence than a lower score.
Does Cortex give me my Turnitin score?
No. Cortex has no connection to Turnitin. Its detector uses GPTZero when configured, which is a separate model reporting a different measurement.
Sources
These were read for this revision. A provider's description of its own product is an attributed claim, not independent certification. Check the current documentation before making a purchase or an academic decision.