Direct answer
Checking text for plagiarism and AI means running one document through two separate detectors that answer two different questions: a similarity check that measures how much of your wording matches existing sources, and an AI-writing check that estimates how much of your prose was machine-generated [1]. Turnitin reports both on the same submission, so a single upload returns a similarity percentage with highlighted matches and an AI score with flagged passages [1]. Neither number is a verdict on your integrity — they are signals you read together, before your instructor reads them.
What Does a Combined Plagiarism and AI Check Actually Measure in Your Text?
A similarity check measures text matching, not wrongdoing. Turnitin compares your submission against its comparison database and reports the percentage of your document that overlaps with those sources, then breaks the overlap down by source so you can see exactly which sentence matched what [2]. Crucially, a correctly quoted and cited passage still counts as a match — the similarity score does not know whether you attributed the words, which is why a well-referenced literature review can legitimately sit in the twenties or higher [2].
The match overview is where the real information lives. Turnitin separates quoted material, cited material, and uncited matches, so a score driven by properly cited quotations reads very differently from a score driven by uncited paragraphs [2]. When you review your own report, sort by the largest uncited match first; that is the block an instructor will open before anything else [2].
The AI check measures something entirely different: statistical patterns in how your sentences are built. Turnitin's AI writing report returns a percentage calculated over qualifying prose only, excluding reference lists, bibliographies, and short quoted fragments from the denominator [1]. It also highlights the specific segments it flagged, so you can see whether the concern is one paragraph or the whole essay [1].
Because the two reports come from the same submission, they share a submission ID and a timestamp, which is what makes them usable as a pair: you can trace any flagged sentence in the AI report back to its position in the similarity report and judge the two signals side by side [1][2].
Why Can a Plagiarism Score and an AI Score Disagree on the Same Document?
The two detectors are not measuring the same property of your text, so disagreement is expected rather than contradictory. Similarity asks "does this wording already exist somewhere in the comparison database?", while AI detection asks "does this wording look statistically machine-produced?" [3]. A document can be 0% matched and still flag heavily as AI, because original machine-written prose matches nothing yet carries the patterns the detector looks for [3].
The reverse also happens. A heavily quoted essay can show a high similarity score with a low AI score, because the matched text is human-authored source material that the AI model does not recognise as machine output [3]. Neither combination tells you the writer cheated; each tells you which detector reacted and why [3].
False positives cluster in predictable places. Highly formulaic academic registers — methods sections, definitions, standard transition phrases, and templated conclusions — read as machine-like because human academics write formulaically too, and the same is true for writers working in a second language whose phrasing is more conventional than idiomatic [3]. If your AI score is concentrated in those sections rather than spread across your argument, that pattern is worth noting in your own revision notes [3].
One practical consequence: never treat a low similarity score as proof you are safe on AI, or a high AI score as proof you plagiarised. Read the two reports as separate diagnostic instruments, then act on whichever one actually flagged something [3].
How Do You Check Text for Plagiarism and AI Before You Submit It?
Self-checking a draft is a legitimate, widely accepted step, and the important variable is which checking route you use. Institutional submissions are stored in Turnitin's student paper database, so a draft submitted through a course assignment link can itself become a future match for other students [4]. A pre-submission check should therefore be non-repository: the file is scanned, reports are returned to you, and the text is not added to the database or shared onward [4].
Prepare the file the way your instructor will receive it. Use .docx, .pdf, or .txt, keep the document in English, and make sure it is the real draft rather than an outline, since both detectors behave differently on short or fragmentary text [1][4]. Keep your revision history and earlier drafts — if a score is ever questioned, dated drafts are the strongest evidence of your own authorship process [4].
Then read the reports in order. Start with the similarity report's uncited matches, fix any genuine attribution gaps, and only then look at the AI report to see which segments were flagged [2][1]. If the AI flags are concentrated in formulaic passages, revise those passages in your own voice rather than rewriting the whole document, and keep the citations and headings intact [3].
Finally, re-check after revision rather than assuming a fix worked. Detectors respond to wording changes, so a second pass on the revised draft is the only way to confirm the flagged segments actually cleared [1][4].
Reading both reports is only useful if the reports are the ones your instructor will actually see. That is exactly what turnitin0 delivers: upload your .docx, .pdf, or .txt and you receive two downloadable PDFs from a single check — a Turnitin AI detection report and a similarity report, matching what appears in institutional systems — usually in under 15 minutes, with the file kept out of the student paper database. With 100,000+ reports delivered and a 4.9/5.0 student rating, it is a way to see the two scores before they matter.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does a high similarity score mean I plagiarised?
No. The similarity score is a text-matching percentage, and correctly quoted, cited material counts as a match [2]. Read the match overview to see whether the overlap is attributed before drawing any conclusion [2].
Can text be 0% similar and still be flagged as AI?
Yes. Original machine-written prose matches no existing source, so the similarity score can be near zero while the AI report still flags segments [3]. The two detectors measure different properties of the same text [3].
Why does Turnitin show *% instead of an AI percentage?
Turnitin displays *% when AI detection falls below its 20% confidence threshold, meaning the signal is low-confidence rather than zero [1]. Treat it as "not enough evidence to score," not as a clean result.
Will checking my draft add it to Turnitin's student paper database?
Not with a non-repository check. Institutional submissions are stored and can match later papers, so a pre-submission check should scan the file without adding it to the database or sharing reports onward [4].
Should I fix the plagiarism score or the AI score first?
Fix attribution first. Genuine uncited matches are a concrete, correctable problem, whereas AI flags often cluster in formulaic passages that need rewriting in your own voice rather than wholesale replacement [2][3].