Direct answer
Turnitin's AI writing detection does not scan for plagiarism — it scans for statistical traces of machine authorship. The detector, built on a transformer model, analyzes two core linguistic signals: perplexity (how predictable a piece of text is) and burstiness (how uniform that predictability is from sentence to sentence) [1]. When both signals match what large language models typically produce, the tool assigns an AI score that appears beside the similarity score in the report [1]. Understanding these two concepts is the key to reading any Turnitin AI writing report with confidence.
What signals does Turnitin's AI writing detector actually use to identify AI-generated text?
Perplexity measures how "surprised" a language model is by a given piece of text. Human writing tends to make less predictable word choices, so it usually scores higher on perplexity; AI-generated text follows the statistically most probable next word, so it scores lower [2]. Turnitin's model compares each segment of your document against what an AI model would most likely produce, and unusually low perplexity is a strong signal of machine authorship [2].
Burstiness describes how much a text's predictability varies across sentences. Human writers naturally mix short, punchy sentences with long, complex ones, which produces high burstiness; large language models tend to stay inside a narrow band of variation for the whole document [2]. When a text shows consistently low perplexity from sentence to sentence, burstiness is low as well, and the prose reads as statistically uniform — a classic fingerprint of AI output [2].
Turnitin combines these two signals inside an AI model trained on academic writing, which lets it distinguish a student's natural prose from text generated by ChatGPT, Claude, or other major LLMs [2]. The detector then returns a percentage reflecting how much of the document appears machine-written. That is why a single paragraph can be flagged even when the overall document is not [2].
How accurate is Turnitin's AI writing detection, and can it flag human writing by mistake?
Turnitin's detector is designed to report only when it is highly confident, yet it is not infallible. The company explicitly advises that the AI report should never be used on its own as proof of misconduct, and that false positives are possible — especially on short documents or heavily edited text [3]. In practice, the shorter the submission, the less reliable the classification becomes [3].
One important accuracy detail is how scores are displayed: any AI score below 20% is shown as an asterisk (*) rather than a single-digit number, so the only explicit low numeric score a student typically sees is 0% [3]. This conservative display reflects Turnitin's own guidance that low scores should be interpreted with caution rather than treated as a guarantee. Instructors are also told to pair the report with their own judgment and a conversation with the student [3].
For students, the practical takeaway is that checking your own draft before submission is the only reliable way to see where the report flags AI-like text. If the detector marks a passage you wrote yourself, that is a signal to revise it — not proof of wrongdoing [3]. Knowing the tool's limits helps you respond to a single percentage without overreacting [3].
How do I see Turnitin's AI writing report for my own draft before I submit it?
The AI writing report is generated automatically whenever a document is submitted through Turnitin, and it appears alongside the similarity report in the instructor's view [4]. Students usually cannot see this report through their own LMS account, which is exactly why many check their drafts through a Turnitin checking service before submitting [4].
With a service like turnitin0.com, you upload your.docx,.pdf, or.txt file and receive the same two reports professors see in their academic systems: the similarity report and the AI writing report [4]. The AI score tells you what percentage of your document is flagged, and the flagged sentences show you exactly which parts of the text are driving the result, so you can target your revisions [4].
The best time to check is before you submit, while you can still edit. A 0% result means the detector found no machine-written text, while anything in the asterisk range (*) sits below the 20% threshold and is reported conservatively [4].
Seeing your own Turnitin AI writing report changes everything: instead of guessing whether your draft will be flagged, you get the exact score, the flagged sentences, and the similarity breakdown before your professor ever does. That is precisely what turnitin0.com is built for — a real Turnitin AI and similarity check on your draft, delivered in minutes, so you can revise with confidence and submit with clarity.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
What is perplexity in Turnitin AI detection?
Perplexity measures how predictable a piece of text is to a language model. Human writing tends to be less predictable (higher perplexity), while AI text follows the most probable next word (lower perplexity), which is why low perplexity is treated as a signal of machine authorship [2].
What is burstiness in Turnitin AI detection?
Burstiness measures how much a text's predictability varies across sentences. Human writing mixes short and long sentences (high burstiness), whereas AI output stays uniformly predictable (low burstiness), making it another key fingerprint the detector looks for [2].
Can Turnitin's AI detector flag human-written text?
Yes, false positives are possible, which is why Turnitin advises instructors never to use the report alone as proof of misconduct and to weigh it with their own judgment [3].
What does the asterisk (*) AI score mean?
Any AI score below 20% is displayed as * instead of a single-digit number; the only explicit low numeric outcome students typically see is 0%. This conservative display is deliberate so low scores are not over-interpreted [3].
Can I check my own paper for AI writing before submitting?
Yes — services like turnitin0.com run your draft through the real Turnitin detector and return the same similarity and AI writing reports professors see, so you know your score before you submit [4].