Direct answer
An AI generator check for plagiarism is not one test but two: a similarity check that compares your text against published sources and a student-paper database, and an AI-writing check that estimates how much of the prose was machine-generated. AI-generated text can score 0% on similarity and still be flagged as AI-written, because those are separate signals measuring different things [1]. Most institutions treat the AI indicator as evidence to discuss, not an automatic misconduct finding, which is why the practical question is what your own draft scores before you submit it [1].
Does AI-Generated Content Count As Plagiarism, And What Does An AI Generator Check Actually Detect?
Whether AI-generated content counts as plagiarism depends on your institution's policy, not on the checker. Similarity detection and AI detection are built on different mechanisms: similarity matching compares your submission against an indexed corpus of web pages, journals, and previously submitted student papers, while AI detection analyses statistical properties of the writing itself — patterns of word choice and sentence construction that language models tend to produce [2]. A passage can therefore be entirely original in the similarity sense and still carry a high AI-writing percentage.
That distinction explains a common confusion. Students run a plagiarism checker, see a low similarity score, and assume they are safe, when the AI indicator on the same report may be far higher [2]. The two numbers answer different questions: "did this text come from somewhere else?" versus "does this text look machine-generated?"
Institutional definitions vary widely, and that variation is the real variable. Some faculties treat unacknowledged AI-generated prose as a form of academic misconduct analogous to ghostwriting; others permit AI assistance with disclosure; others restrict it to specific tasks [2]. Because the definition is local, the same draft can be acceptable in one module and a violation in another.
The practical takeaway is that no checker — Turnitin's included — issues a verdict. Detection output is designed as one input into an academic judgement, and institutions are repeatedly advised to interpret indicators in context rather than as proof [2].
How Does Turnitin's AI Writing Detection Work Alongside Its Similarity Check?
Turnitin's AI writing detection runs inside the same report workflow as the similarity check, which is why students often conflate them. The AI Writing Report presents an overall percentage of qualifying prose estimated to be AI-generated, together with a highlighted overlay showing which sentences triggered the indicator, so a reader can see exactly where the signal sits rather than just a headline number [3].
The detection is deliberately scoped. It processes long-form, English-language prose and requires a minimum amount of text before it will produce a result, and it excludes elements that are not the author's own continuous writing — quotations, reference lists, and short fragments are not counted in the same way as body text [3]. That scoping matters when you are trying to interpret a score on a mixed document such as a lab report with tables and citations.
Accuracy claims deserve careful reading. Turnitin positions the indicator as a probability estimate with a low false-positive rate on its validation data, but it explicitly acknowledges that formulaic, highly structured, or heavily edited writing can produce misleading signals [3]. Non-native English writers and students working in tightly templated genres are the populations most often discussed in that context, and Turnitin recommends human review rather than automatic penalties [3].
Because the AI indicator and the similarity index appear side by side, the two are frequently read as one composite score. They are not. A report can show a clean similarity index and a substantial AI percentage, or the reverse, and each number needs to be read in its own terms [3].
Can Students Check AI-Generated Text For Plagiarism And AI Score Before Submitting?
Students generally cannot run the check that matters most — the one attached to their actual assignment — because that check is controlled by the instructor and the institution. Turnitin's own student-facing guidance explains that whether a student can submit a draft and view a report depends on whether the institution has enabled that option for the assignment; where it is not enabled, there is no self-service route inside the institutional account [4].
That gap is why pre-submission practice checks exist as a category. The function students are actually looking for is a non-repository check: upload a draft, receive the same two reports an instructor would see — AI writing and similarity — without the file being added to the institutional student-paper database and without the report being shared with the instructor [4]. The "non-repository" detail is the one that matters ethically, because a draft that enters a paper database can later be matched against your own final submission.
When you do run a practice check, read it the way an instructor would. Look at the AI overlay to see which sentences triggered the indicator, and check whether those sentences are ones you actually wrote or ones you imported from a generator and lightly edited [4]. Then look at the similarity matches individually: a citation formatted identically to a source is a different problem from a copied paragraph.
The most useful outcome of a pre-submission check is not a number but a decision. If the AI indicator concentrates in a few passages, rewriting those passages in your own reasoning — rather than paraphrasing the same structure — is what changes the underlying signal [4]. If it is spread evenly across a long document, the drafting process itself is the issue, and that is a conversation to have before submission rather than after.
Reading your own report before your instructor does turns an anxious guess into a specific edit list — and turnitin0 exists to make that possible, returning the same AI and similarity reports your university sees, typically in under 15 minutes, without adding your draft to any student-paper database.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is AI-generated text always flagged as plagiarism?
No. Plagiarism detection measures overlap with existing sources, and fully generated text often has little of that. The AI-writing indicator is a separate estimate and may be high even when similarity is near zero [1].
Can a plagiarism checker detect that I used ChatGPT?
A similarity checker alone cannot. Turnitin's AI writing detection analyses patterns in the prose and reports a percentage with a sentence-level overlay, which is a different output from a similarity match [3].
Why does Turnitin show *% instead of a number?
An asterisk appears when the AI-writing result falls below Turnitin's confidence threshold, so the report signals a low-confidence estimate rather than a precise figure [1].
Does checking my draft first count as misconduct?
Running a practice check on your own work is not itself a policy violation, but your institution's rules on AI assistance are what govern the draft itself, so read the module policy before relying on any score [2].
Will my instructor see a pre-submission check?
Not from a non-repository practice check, which is not linked to your institutional assignment and is not shared with your instructor [4].