Which AI Detector Does Turnitin Use?
Turnitin runs its own proprietary AI-writing detector, not a third-party tool like GPTZero or Originality.ai. Here's what it checks, how it reports, and why students never see the score.
Turnitin uses its own proprietary, in-house AI writing detector, which it built and launched in 2023 — not a third-party service like GPTZero or Originality.ai. It produces a separate AI-writing indicator, a percentage estimate of how much text appears AI-generated, that is shown alongside but calculated independently from the Similarity (plagiarism) score.
Turnitin uses its own detector
Turnitin's AI writing detection is a proprietary, in-house model that Turnitin developed itself and launched in April 2023. It is not a rebranded or licensed version of another company's tool — Turnitin built and trains its own classifier on large sets of paired human and AI-written text, drawing on the student-writing data already flowing through its plagiarism-checking service.
Since launch, Turnitin has expanded from a single model to a combination of models designed to flag directly AI-generated text, AI text that has been paraphrased, and AI text that has been altered by so-called 'humanizer' tools. The company reports the detector has been used to review well over 200 million papers across thousands of institutions.
How Turnitin's AI indicator works
Turnitin's model reads long-form prose and scores it at the sentence level, giving each sentence a value between 0 (confidently human) and 1 (likely AI-generated). To do this it examines the document in overlapping segments of roughly 250 words, then aggregates those sentence-level scores into a single overall percentage.
That percentage is an estimate of how much of the document Turnitin's software believes was generated by AI writing tools — not a definitive verdict. Turnitin itself frames the number as a prompt for human review rather than proof of misconduct, and independent studies have documented meaningful false-positive rates, including higher flag rates for non-native English writers and very formulaic academic prose. A percentage is a signal, not a conclusion.
Does Turnitin use GPTZero or Originality.ai? (No)
No. GPTZero, Originality.ai, Copyleaks, and Turnitin are separate, competing products, each with its own independently built detection model. Turnitin does not call out to GPTZero or Originality.ai behind the scenes, and those tools do not feed into Turnitin's score.
This is why the same piece of writing can receive different AI percentages from different detectors: each vendor trains on different data and calibrates its thresholds differently. Turnitin's model is tuned specifically for academic writing and is generally described as conservative, but it remains its own system with its own strengths and blind spots.
Similarity vs AI-writing: two different reports
It's easy to confuse the two scores, but they measure completely different things. The Similarity Report checks your text against a database of published work, web pages, and other student submissions to find matching or unoriginal passages — that's plagiarism detection. The AI writing indicator instead estimates how much of the text reads as machine-generated, regardless of whether it's original.
The two run independently: AI-writing highlights don't appear in the Similarity Report, and a low similarity score tells you nothing about the AI indicator. A paper can be 0% similar (fully original) yet still receive a high AI-writing percentage, or the reverse. Knowing the difference matters when you're interpreting the feedback.
Why students never see the score (institution-only)
Turnitin's AI writing detection is an institution-controlled feature. It's enabled and administered by schools and instructors, and in most setups the AI Writing Report is hidden from the student view entirely. Students generally cannot log in and run their own paper through Turnitin's AI detector on demand.
That means the first time a student learns their AI-writing percentage is often after submission, from an instructor who chooses to share it. There is no official student-facing self-service portal for the AI indicator — access sits with the institution, at the educator's discretion.
Where humantext.pro fits
Because students can't run Turnitin's AI detector themselves, it helps to understand how your own draft reads before you hand it in. humantext.pro offers a free, no-signup AI-writing self-check: paste your text and see which passages sound machine-generated, so you can revise robotic or repetitive phrasing into clearer, more natural writing in your own voice. To be clear about scope, this checks the AI-writing signal only — it is not a plagiarism database and doesn't reproduce Turnitin's Similarity score, so pair it with a dedicated plagiarism checker if you also need an originality check. Think of it as a pre-submission sanity check and a writing-improvement tool, not a stand-in for Turnitin itself.
Turnitin AI Detection — FAQ
Does Turnitin use GPTZero?
No. Turnitin uses its own proprietary AI writing detector, built and maintained in-house. GPTZero is a separate, competing company with its own model, and the two do not share results or integrate with each other.
Is Turnitin's AI detector accurate?
Turnitin advertises high accuracy (around 98%, with a low false-positive rate for documents above a certain AI threshold), but independent research has found notably higher false-positive rates in practice, especially for non-native English writers and formulaic prose. Turnitin itself says the score should prompt human review, not serve as standalone proof.
Can I check my writing before Turnitin sees it?
Not through Turnitin directly — it's institution-controlled and typically unavailable to students. You can, however, run your own pre-submission self-check with a free tool like humantext.pro to see which passages read as AI-generated and revise them into clearer, more natural writing before you submit. It checks the AI-writing signal only, not plagiarism.
What's the difference between the Similarity score and the AI writing indicator?
The Similarity score measures overlap with existing sources (plagiarism). The AI writing indicator estimates how much text appears machine-generated. They're calculated independently and reported separately — a paper can score low on one and high on the other.
When did Turnitin launch its AI detector, and what does it detect?
Turnitin launched AI writing detection in April 2023 and has updated its models since, adding coverage for paraphrased text and text altered by 'humanizer' tools. It's designed to recognize output from a range of modern language models, and the underlying models continue to be refined over time.
