Direct answer
Checking text for AI generation means running it through a detector that scores how much of the prose matches patterns typical of large language models, then reading that score as evidence rather than a verdict. Turnitin's AI Writing Report is the detector most university students will meet, and it returns an overall percentage of "qualifying text" plus sentence-level highlights showing exactly which passages triggered the flag [1]. The number is only useful if you know what it counts, what it ignores, and where it can be wrong — which is what this guide covers.
How Do You Check Whether Text Was Generated by AI?
You check text for AI generation by submitting it to a detector built for that purpose and reading the resulting report, not by looking for tell-tale words or "AI-sounding" phrasing. Turnitin's detector analyses word choice and sentence structure across a document and estimates the share of qualifying text that appears machine-generated, then marks the specific segments it flagged so a human can inspect them [2]. That structure matters: the report is designed to point a reader at passages, not to hand down a judgement about the writer.
Detection is also narrower than most people assume. Turnitin's own guidance is explicit that the tool looks for AI-generated text and is not a reliable measure of AI-assisted editing, so a paragraph you drafted yourself and then had ChatGPT tighten can behave unpredictably in either direction [2]. Length is a further constraint — very short pieces give the model too little signal, and Turnitin's institutional guidance treats roughly 300 words as the practical floor for meaningful detection, which is why a 150-word discussion post often returns a low-confidence result rather than a clean score [2].
The practical consequence is that no single check should be treated as final. Detectors can flag human writing and can miss machine writing, and Turnitin tells institutions as much in its own documentation [2]. If you are checking a document that matters — a dissertation chapter, a grad school application essay, a client deliverable — the sensible sequence is to run the check, read the highlighted passages yourself, and compare them against your own drafting history before drawing any conclusion.
What Does the AI Percentage in a Detection Report Actually Mean?
The percentage in an AI detection report is the proportion of qualifying text that the detector believes was AI-generated — not the proportion of your whole document [3]. Turnitin excludes blocks it cannot assess, such as reference lists, quotations, and very short fragments, from the qualifying-text calculation. So a report showing 40% on a 4,000-word essay does not mean 1,600 words were flagged; it means 40% of the portion the detector considered assessable crossed its threshold [3].
That distinction explains the two results students most often misread. A 0% is not a certificate of human authorship — it means nothing in the qualifying text crossed the confidence threshold, which is a different claim entirely [3]. And a high score does not identify intent: the report shows which passages look machine-generated, and it is the reader's job to work out why, whether that is a pasted AI draft, a heavily templated section, or a false positive on formulaic academic prose [3].
Turnitin also withholds a precise figure when its confidence is low, displaying an asterisked value instead of a hard number [1]. That display is a signal in itself: it tells you the detector found too little evidence to commit, so treating the result as a pass or a fail is a misreading. When you see it, the useful move is to look at the segment highlights and the word count of the qualifying text, then decide whether the document needs another pass or simply more of your own writing in the flagged sections [1].
Can You Check Your Own Draft With Turnitin Before You Submit It?
In most institutions, students do not have direct access to Turnitin's AI detector — submissions run through the class assignment, and the report goes to the instructor rather than the student [4]. That is a deliberate design choice: Turnitin positions the AI indicator as one piece of evidence an instructor weighs alongside drafts, citation practice, and the student's own account of how the work was produced, not as a standalone accusation [4]. Students who want certainty before a deadline therefore usually cannot get it from the institutional route.
What students can do is reduce the risk that a report surprises them. Keep versioned drafts, keep your notes and source material, and be able to explain the provenance of any section that reads as unusually polished [4]. If you did use a model for brainstorming or restructuring, saying so is generally treated more leniently than a flagged passage you cannot account for — Turnitin's own academic-integrity guidance frames transparency about AI use as the safer position than guessing what the detector will do [4].
The gap this leaves is a preview: a way to see the same AI and similarity report your instructor would see, on your own draft, while you can still revise. That is precisely the use case an independent pre-submission checker is built for, and it is worth understanding how the report you receive maps onto the one that lands in your instructor's LMS before you rely on either.
If the uncertainty above is the part that is costing you sleep, turnitin0 closes it: upload your .docx, .pdf, or .txt (English, 300–30,000 words, under 20 MB) and you get back the two PDFs that matter — a Turnitin AI detection report and a similarity report, matching what your professor sees — usually in under 15 minutes, without your file entering Turnitin's student paper database. Over 100,000 reports have been delivered to more than 20,000 students across the US, UK, Canada, Australia, New Zealand, and Ireland, at $3.80 per check with no subscription. Read the score while you can still edit the paragraph that caused it.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
Get Real Turnitin AI & [Similarity Report](https://www.turnitin0.com/guides/us/turn-in-ai-detector)
FAQ
Is there a free way to check text for AI generation?
Free detectors exist, but they generally do not use the same model as Turnitin and will not reproduce the score your instructor sees. If the result has consequences, check with the detector your institution actually runs [1].
Can a Turnitin AI score be wrong?
Yes. Turnitin documents both false positives and missed AI text, which is why the report highlights segments for human review instead of issuing a verdict [2].
Does a 0% AI score prove the text is human-written?
No. It means no qualifying text crossed the confidence threshold — a narrower claim than human authorship [3].
Why does my report show an asterisk instead of a number?
Turnitin displays an asterisked value when its confidence is below its threshold, signalling a low-confidence result rather than a definitive score [1].
Should I tell my instructor I used AI to help draft?
Turnitin's academic-integrity guidance treats transparency about AI use as the safer position, since the indicator is only one of several things an instructor weighs [4].