Direct answer
Testing for AI and plagiarism means running two separate checks on the same document: an AI-writing detector that estimates how much of your prose looks machine-generated, and a similarity check that matches your text against student papers, journals, and web sources [1]. Turnitin produces both as two distinct reports from a single submission, and they measure different things — a clean similarity score says nothing about the AI indicator, and an AI score of 0% does not prove the words are yours [2]. The practical move before a deadline is to run both checks on your own draft, read the two reports the way your instructor will, and fix the flagged passages while you still can.
What is the difference between an AI detection test and a plagiarism check?
A plagiarism check is a matching exercise. Turnitin compares your document against its repository of previously submitted student papers, academic journals, and indexed web pages, then reports the percentage of your text that overlaps with those sources alongside a numbered list of where each match came from [2]. A hit is evidence of textual overlap, not a verdict — correctly quoted and cited material still appears as a match, which is why the similarity index is designed to be read with its source list rather than on its own [2].
AI detection works on a different signal entirely. Instead of looking outward for matching sources, it looks inward at the statistical properties of your sentences — the predictability and uniformity of word choice that tend to characterise text produced by large language models [1]. Turnitin's detector only evaluates prose that qualifies as long-form writing, which is why very short documents or heavily bulleted pages can return no meaningful AI figure at all [1]. Nothing in that process checks whether your ideas came from someone else.
The two checks therefore fail in opposite directions, and that is the reason you need both. Text you wrote entirely yourself can score high on similarity because you paraphrased a source too closely, while text you generated with ChatGPT and then lightly edited can score 0% similarity because it matches nothing in any database [2]. Relying on one number to answer both questions is the single most common misreading students make before submission.
Can you test a draft for AI and plagiarism before submitting it to Turnitin?
Yes — but the route matters, because a draft check is only useful if the report matches what your instructor will see. If your module allows draft submissions, Turnitin will generate a Similarity Report for a draft you submit to the assignment, and your instructor controls whether students can view similarity scores and whether drafts are permitted at all [3]. Where draft submissions are switched off, or where you simply do not want your file added to the institutional paper database, an independent pre-submission check is the workable alternative [3].
The detail that decides whether a draft check is honest is the repository setting. A check that stores your file in a student paper database can inflate later similarity results and creates a record you may not control, so look for a non-repository check that returns the report without retaining the document in a shared database [3]. Equally, the report itself has to be the real Turnitin output — the same AI indicator and the same similarity layout your marker opens — because a paraphrased "AI likelihood" percentage from a generic detector tells you nothing about the number that will actually be attached to your work [1].
Timing is the other variable worth planning around. Both checks need your document in its near-final form, since the AI indicator shifts as you edit and the similarity index shifts as you add or cut quotations [1][3]. Running the test on a 60% draft and then rewriting two sections leaves you testing a document that no longer exists.
What do the Turnitin AI score and similarity index actually mean on the report?
On the AI side, the report shows the percentage of your qualifying prose that the detector classifies as AI-generated, and Turnitin deliberately withholds precision at the low end: when the signal falls below its confidence threshold, the report displays an asterisk in place of an exact figure, signalling a low-confidence result rather than a clean bill of health [4]. That asterisk is not a pass. It means the detector found some signal but not enough to state a percentage, and it is the band most students misread as "0%" [4].
The similarity index is a percentage of matched text, and it is not a plagiarism score. Turnitin's own guidance is that the number should be read through the match list, because a single source contributing a long block of matched text means something different from twelve sources each contributing a correctly cited phrase [4]. Quotations and bibliography entries are typically excluded from the index, so a high figure usually points at paraphrasing that stayed too close to the original wording [2][4].
Read together, the two figures tell you which problem you have. High similarity with a low AI indicator points to citation and paraphrasing work; a low similarity index with a flagged AI percentage points to authorship, and rewriting for source attribution will not move it [1][4]. The reports also carry the segment-level highlighting that shows exactly which paragraphs triggered each signal, which is the part worth acting on rather than the headline number.
Reading your own two reports is straightforward once you have them — the hard part is getting a genuine pair before the deadline rather than guessing from a third-party approximation. That is exactly what turnitin0 was built for: upload your draft and receive the Turnitin AI detection report and the similarity report together, in the form your instructor will open, so the paragraph-level flags you act on are the real ones.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does Turnitin check AI and plagiarism in the same submission?
Yes. One submission generates two separate reports — an AI writing report and a similarity report — and they are read independently [1][2]. A single upload is enough to see both, but the numbers answer different questions.
Is a 0% similarity score proof that I did not use AI?
No. Similarity measures overlap with existing sources, and AI-generated text often matches nothing in any database, so it can score 0% similarity while still being flagged by the AI detector [2][4]. The two indicators have to be read side by side.
What does the asterisk on the Turnitin AI score mean?
It means the AI signal fell below Turnitin's confidence threshold, so the report shows an asterisk instead of an exact percentage [4]. Treat it as a low-confidence signal worth reviewing at paragraph level, not as a confirmed zero.
Can I test my draft without adding it to my university's paper database?
That depends on how the check is configured. Institutional draft submissions may store the file, whereas a non-repository check returns the reports without adding your document to a shared student paper database [3]. Check the setting before you upload.
How close to the deadline should I run the check?
Run it on the version you intend to submit, and leave enough time to revise — the AI indicator moves as you rewrite and the similarity index moves as you adjust quotations [1][3]. A check on an early draft tells you about a document you are no longer submitting.