Direct answer
An AI content detector is a classifier that estimates whether a piece of text was written by a large language model rather than a person, and it does so by measuring how statistically predictable the writing is rather than by searching for a hidden watermark. Turnitin's detector splits a submitted document into sentences and flags each one, then reports the share of sentences it believes were AI-generated [1]. Those results are probability signals, not proof of misconduct, and Turnitin itself advises that they be read alongside context [3].
How Does an AI Content Detector Actually Decide That Text Was Written by AI?
Detectors do not read for meaning, style, or "soul". They measure predictability: how likely each next word is given everything before it, and how much that predictability fluctuates across a paragraph [2]. Human academic writing tends to swing between long, clause-heavy sentences and short, blunt ones, while model output often settles into a narrow, even rhythm. That evenness is what a classifier latches onto, which is why a technically flawless paragraph can score as AI while a slightly clumsy human paragraph scores as human [2].
Turnitin's implementation is deliberately narrow. It was trained on academic and web text and is designed for longer-form student submissions, which is why a report is only generated for documents above roughly 300 words and in English [1]. The report does not colour the whole document one shade; it highlights individual sentences it classifies as AI-generated so a reader can see exactly which lines triggered the score [1]. That sentence-level view is the single most useful feature of the report, because it separates a genuinely flagged passage from a document that merely contains a few predictable sentences.
It is also worth being precise about what the number means. The headline percentage is the proportion of qualifying sentences flagged as AI, not a confidence rating and not a plagiarism figure [1]. A 30% AI score does not mean "30% guilty"; it means roughly three in ten sentences matched the classifier's pattern for machine text. Because the underlying method is statistical, the output is a probability estimate that can be wrong in both directions, and Turnitin positions it as a signal for human judgement rather than a verdict [2].
Why Does Turnitin Show an Asterisk (*%) Instead of a Percentage in the AI Writing Report?
An asterisk appears when the detected AI share falls below Turnitin's 20% confidence threshold. In that band the system is telling you it is not confident enough to publish a firm number, so it displays *% instead of, say, 11% [3]. This is the detail most students misread. *% is not a zero, and it is not an error — it is a low-confidence signal, and it is the most common result for drafts that were written by hand and then lightly polished with a model.
The threshold exists because short, formulaic, or heavily templated prose is genuinely hard to classify. Reference lists, methods boilerplate, definitions copied from a textbook, and sentences that follow a required structure all look statistically predictable without being machine-written [3]. Turnitin also limits report generation to English documents over 300 words, so a short abstract or a bulleted outline may produce no AI report at all [3]. If you are staring at a report and cannot find a percentage, check the word count and the language before assuming the worst.
The practical consequence is that *% should be read as "no strong AI signal detected", not as "clean" and certainly not as "flagged". Institutions that use Turnitin are advised to treat the score as one input among several and to weigh it against drafting history, version control, and a conversation with the student [3]. For a student, the useful move is to look at which sentences are highlighted rather than fixating on the headline figure — a low-confidence overall score with two flagged sentences is a very different situation from a high score spread across the whole document.
Can You Check Your Own Draft With a Real Turnitin AI Report Before You Submit It?
Not through Turnitin directly. Student access to a Turnitin report is normally granted only when an instructor enables it for a specific assignment, so there is no general self-service channel for checking a draft outside a class [4]. That gap is exactly why third-party pre-submission checks exist: they run the same style of AI and similarity analysis and return the report to you, before anything reaches your instructor's inbox.
The detail that matters most for a draft you intend to submit is repository status. A standard institutional submission can be stored in the student paper database, which creates a genuine risk if you later resubmit revised work and it matches your own earlier upload. A non-repository check avoids that by not adding the file to the database and not sharing reports with third-party databases [4]. When you compare AI content detectors, ask two questions: does it return both the AI writing report and the similarity report, and does it keep the file out of the paper repository [4]?
Finally, treat a pre-submission report as diagnostic rather than decorative. The value is in the sentence-level flags and the similarity matches, which tell you precisely which paragraphs need rewriting, re-citing, or a clearer human voice [4]. Students who read the highlights and revise accordingly get far more out of a check than students who only look at the top-line number and hope it moves.
If you want that same sentence-level view of your own draft — the AI score, the flagged sentences, and the similarity summary — before your deadline rather than after it, turnitin0 runs the check for you and hands back the reports. Upload a.docx,.pdf, or.txt file over 300 words, and most orders are delivered in under 15 minutes, with the file kept out of the student paper database.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is an AI content detector accurate enough to accuse someone?
No. Turnitin describes its AI writing indicator as a probability-based signal and advises institutions to interpret it with context rather than as standalone evidence [2]. A flagged sentence means the text matched a statistical pattern, not that misconduct occurred.
What does *% mean on a Turnitin AI report?
It means the detected AI share fell below the 20% confidence threshold, so Turnitin displays an asterisk instead of a specific percentage [3]. It is a low-confidence result, not a zero and not a failure.
Why did my report show no AI percentage at all?
AI reports are generated only for English documents above roughly 300 words [3]. Short files, non-English text, and some file types will not produce an AI indicator even when a similarity report is available.
Can I lower a score I have already been given?
You can revise the flagged sentences in your own voice and re-check the draft before submitting. The useful target is the highlighted passage, not the headline number [1].
Does checking my draft with a third-party service affect my institution's Turnitin account?
A non-repository check does not add your file to the student paper database and does not share reports with third-party databases [4]. Your institutional submission remains a separate process controlled by your instructor.