Direct answer
You check whether text was created by AI by running it through an AI detector that measures statistical patterns in the writing, then reading the resulting report rather than the headline number alone. Turnitin's AI writing detection identifies text likely generated by large language models and reports the percentage of qualifying text flagged, with per-segment highlighting so you can see exactly which passages triggered it [1]. Detection is a signal about how the text reads, not proof of who wrote it, and Turnitin itself treats results as a prompt for review rather than a verdict [1]. The most reliable workflow combines a detector report with your own reading of the flagged passages, because no tool can confirm authorship or intent on its own [1].
How do AI detectors actually decide whether text was written by AI?
AI detectors work by measuring statistical properties of language rather than by searching for a hidden watermark. The two signals that matter most are perplexity, which captures how predictable each word choice is given the words before it, and burstiness, which captures how much sentence length and structure vary across a passage [2]. Text produced by large language models tends to sit at low perplexity and low burstiness — fluent, evenly paced, and statistically unsurprising — which is precisely the pattern detectors are trained to recognise [2].
Turnitin's approach splits a submission into overlapping segments and evaluates each window separately, then reports the proportion of qualifying text that looks machine-generated [2]. That segment-level design is why a report can flag three paragraphs and leave the rest clean, and why the resulting figure is a proportion of analysed text rather than a probability about the document as a whole [2]. Only text that meets Turnitin's minimum length and formatting criteria is analysed at all, so very short submissions may return no score [1].
Because the output is probabilistic, several practical consequences follow. Human editing layered over an AI draft can lower a flag without removing it entirely, since rewriting a few sentences may not change the underlying statistical profile of the surrounding passage [2]. Conversely, formulaic academic writing — heavy on stock transitions and uniform sentence length — can read as low-burstiness even when a person wrote every word [1]. The honest takeaway is that detectors estimate how a passage reads, and that estimate needs a human reader to interpret it [1][2].
What does a Turnitin AI writing score really tell you about a document?
The AI writing score is the percentage of qualifying text that the detector flagged as AI-generated — it is not the probability that the whole document was written by AI [3]. If a 4,000-word essay returns 18%, that means roughly 18% of the analysed prose met the flagging criteria, not that there is an 18% chance a chatbot produced the paper [3]. Reading the score without opening the highlighted segments is the single most common way students and instructors misread a report [3].
Turnitin also handles low-confidence results differently from confident ones. When AI detection falls below its 20% confidence threshold, the report displays *% instead of an exact percentage, signalling that the signal is weak rather than clean [3]. Scores are only generated for submissions above a minimum word count, and only qualifying prose is analysed, so quotations, reference lists, and short-answer items may sit outside the calculation entirely [3].
Institutional guidance is consistent on how the number should be used. A high score is a reason to look at the flagged passages and talk with the writer, not an automatic misconduct finding, because detectors can flag polished human prose and can miss lightly edited machine text [3]. Treat the score as the start of an evidence trail: open the report, read each highlighted segment, and ask whether the writing style there matches the rest of the submission [3][1].
Can you check a draft for AI detection before you submit it?
In most institutions, students cannot run a Turnitin AI check on their own draft inside the university's Turnitin account. The draft folder that would allow self-checking has to be enabled by the instructor, and many courses never turn it on, which is why students often see no way to test a paper before the deadline [4]. That gap is exactly where pre-submission checking services fit, and it is the reason they exist [4].
A third-party pre-submission check returns the AI detection report and the similarity report together, so you can see both the flagged passages and any matching sources while you still have time to revise [4]. The important operational difference is repository status: a check that does not add your file to Turnitin's student paper database avoids the situation where your own draft later matches itself in the institutional run [4]. That matters most for students who intend to revise and resubmit the same document.
Pre-submission checking is a complement to the instructor's final run, never a replacement for it, because the final submission is the one that counts and the settings may differ [4]. Used well, it turns an anxious guess into a concrete revision list: you see which paragraphs read as machine-generated, rewrite them in your own voice, add the citations that were missing, and re-check before the deadline [4][3]. For students working from an AI-assisted draft, that loop is the difference between submitting blind and submitting informed [4].
If you would rather see the actual report than guess at your odds, turnitin0 lets you run the same Turnitin AI and similarity check your instructor will run, on your own draft, before the deadline closes.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Can a single AI detector result prove a document was written by AI?
No. Detectors estimate how text reads using statistical patterns, and Turnitin presents its output as a review signal rather than a finding of authorship or intent [1]. A conclusion about a document should rest on the flagged passages themselves plus context the tool cannot see [3].
Why does Turnitin show *% instead of a number?
Turnitin displays *% when AI detection falls below its 20% confidence threshold, meaning the signal is too weak to report as a precise percentage [3]. It is a low-confidence reading, not a clean bill of health, so the highlighted segments are still worth reading [3].
Does rewriting AI-drafted text always remove the flag?
Not always. Detectors respond to perplexity and burstiness, so light edits that preserve the original sentence rhythm may leave the statistical profile largely unchanged [2]. Deeper rewriting in your own voice, with your own citations and examples, is what actually shifts the reading [2].
Can students check their own work in Turnitin before submitting?
Usually not inside the institutional account, because the draft folder must be enabled by the instructor [4]. Pre-submission checking services fill that gap by returning AI and similarity reports without adding the file to the student paper repository [4].
What should I do if my own writing is flagged?
Open the report, read each highlighted segment, and compare the style there with the rest of the document [3]. Then revise the flagged passages, add any missing citations, and re-check before the deadline so you submit with evidence rather than anxiety [4].