A Turnitin AI score is an estimate of how much eligible prose appears likely to have been generated by AI or modified with an AI text-altering tool. It is a review signal—not a grade, a plagiarism percentage, or proof of misconduct.
The short answer
In current reports, the visible outcomes are not a smooth 0-to-100 scale. A report may show 0%, *%, a number from 20% to 100%, or no result. Turnitin suppresses exact values from 1% to 19% and displays an asterisk because it says false positives occur more often in that low range.
That distinction matters. A “40%” result means the system marked about 40% of the eligible prose, not that there is a 40% chance the whole paper was written by AI. The percentage should lead to a closer look at highlighted passages, the assignment rules, and the writer’s process.
Turnitin AI score breakdown: 0 to 100
Turnitin officially distinguishes 0%, the suppressed 1–19% range, and visible scores from 20–100%. The additional bands below are a practical reading framework—not official Turnitin penalty thresholds.
The system found no eligible prose it classified as likely AI-generated or AI-altered. This does not authenticate authorship; it only describes this scan.
The exact number and highlights are withheld in current reports because this range is less reliable. Treat it cautiously and avoid drawing conclusions from the asterisk alone.
Review the marked passages in context. Check whether they are formulaic prose, legitimate quotations, standard definitions, or sections that do not match the writer’s usual process.
A structured review is sensible: compare drafts and version history, discuss permitted AI use, inspect citations, and ask the writer to explain their reasoning.
This is a strong signal for deeper review, not automatic proof. Confirm what counts as qualifying text and apply the institution’s policy with human judgment.
Important: these practical bands describe review intensity only. Your school’s policy—not the number alone—determines what happens next.
What does the AI score actually measure?
The denominator is not necessarily every word in the file. Turnitin describes “qualifying text” as prose sentences in long-form writing, such as an essay, article, or dissertation. Lists, tables, code, poetry, scripts, and annotated bibliographies may not be evaluated reliably or counted in the same way.
That can produce a result that looks surprising. If a 2,000-word submission contains 1,200 words of qualifying prose and 600 of those words are highlighted, the displayed figure may be around 50%—even though the highlights cover only 30% of the entire file.
Four reasons scores can change
- Text length and format: short or non-prose content may not qualify for a report.
- Editing: genuine human revision can change statistical patterns, while AI paraphrasers can create other detectable patterns.
- Language: support and detection features vary by language.
- Model updates: detection systems evolve, so a result is a snapshot from a particular version and time.
Turnitin’s current guide says an eligible file generally needs at least 300 words of long-form prose, no more than 30,000 words, a supported language, and a supported file type. A gray or unavailable result usually means the file could not be processed—not that it “passed.”
AI writing score vs. Similarity Score
| Report | What it estimates | What it cannot prove |
|---|---|---|
| AI writing | Share of qualifying prose likely generated by AI or altered by an AI text tool | Intent, misconduct, authorship, or plagiarism |
| Similarity | Share of text matching material in compared sources | Whether every match is plagiarism or whether AI was used |
A paper can have a low Similarity Score and a high AI writing score. It can also have a high Similarity Score for acceptable reasons, such as properly quoted passages, references, or assignment templates. Always open the report and inspect the text behind the number.
What to do with a score: a responsible workflow
- Read the assignment’s AI policy.Permitted brainstorming, grammar support, translation, and generative writing differ across courses. The policy supplies the context the detector cannot.
- Inspect the highlighted passages.Look for patterns, but also check quotations, technical phrasing, standard definitions, and other legitimate explanations.
- Review evidence of process.Outlines, notes, source annotations, document history, and earlier drafts can show how the work developed.
- Have a human conversation.Ask the writer to explain an argument, source choice, or revision. Turnitin itself says its model can misidentify text and should not be the sole basis for adverse action.
- Revise for substance, not for a detector.Add specific evidence, clarify reasoning, cite sources correctly, and make sure the prose accurately reflects the writer’s own understanding.
Check your draft before submission
PaperCheck provides a fast, accurate AI-writing estimate at no cost, so students and writers can review suspicious-looking passages before submitting. It is built around data minimization: text is processed in real time and deleted immediately after analysis; results may be cached only for the active session, according to our privacy policy.
Use any detector as a second opinion, not a guarantee. Different tools use different models, so PaperCheck cannot reproduce or promise a particular Turnitin result.
Frequently asked questions
Is 20% AI on Turnitin automatically bad?
No universal score equals misconduct. A visible 20% means about one fifth of qualifying prose was identified as likely AI-written or AI-altered. The marked text, assignment rules, writing process, and human review matter more than a standalone threshold.
Why did I get *% instead of a number?
The asterisk represents a result in the 1–19% range. Turnitin withholds the exact figure and highlights because it reports a higher incidence of false positives in this range.
Can human writing receive a high AI score?
False positives are possible with any probabilistic detector. Formulaic, highly predictable, heavily edited, or unusual prose may contribute to a result. Keep drafts and version history, and ask for a contextual review if your original work is questioned.
Can PaperCheck guarantee the same score as Turnitin?
No. PaperCheck is an independent pre-submission checker and is not affiliated with Turnitin. Because detection models differ, no third-party tool can guarantee an identical score. It is useful for finding areas worth reviewing before submission.
Is PaperCheck really free and data-safe?
The standard checker is free. PaperCheck’s privacy policy states that submitted text is processed in real time and deleted immediately after analysis, with results only potentially cached for the current session. Writers who need an unlimited Pro workflow can use AceEssay.