Direct answer
Checking for AI and plagiarism means running two separate tests on the same draft: a similarity check that compares your text against student papers, web pages and publications, and an AI-writing check that estimates how much of your prose reads as machine-generated. They answer different questions, use different comparison methods, and produce two different reports — which is why a single "score" can never cover both. A draft can be 100% original and still carry a high AI indicator, or be heavily AI-assisted and show a low similarity index [1]. Understanding that split is the whole point of checking before you submit.
What is the difference between an AI detection check and a plagiarism check?
A plagiarism check is a matching exercise. Turnitin's similarity report compares your submitted text against a repository of previously submitted student papers, current and archived web pages, and scholarly publications, then highlights every span that matches a source and links it back to that source [2]. The output is a similarity percentage plus a match list, and the percentage is simply the proportion of your document's text that overlaps with something in that comparison set [2]. Crucially, the number is not a verdict: a high index is often produced by properly quoted material, standard methodology phrasing, or a reference list, so each highlighted span has to be inspected for correct quotation and citation before anyone calls it plagiarism [2].
An AI detection check works differently. It does not look for matching sources at all — it analyses the statistical and stylistic properties of your prose to estimate how likely it is that a language model produced it. Turnitin's AI writing indicator is reported as a separate layer alongside the similarity score and explicitly does not replace it, because the two are generated from entirely different comparison processes [1]. The AI figure represents the share of qualifying long-form English prose that the model flagged, not a probability that you cheated [1].
The practical consequence is that neither report can substitute for the other. A similarity report has no mechanism for spotting text that a model wrote from scratch, and an AI indicator has no mechanism for spotting a paragraph you copied from an unattributed source. Institutions therefore review both, side by side, and expect students to understand the distinction too [1][2]. If you only ever look at one number, you are checking half of your exposure.
There is also an interpretive gap worth naming. Turnitin positions both reports as decision support rather than proof, instructing reviewers to apply academic judgement to the highlighted evidence instead of treating a flag as a finding of misconduct [1]. That framing matters for students: a flagged span is a prompt to look closer, not a conviction — and the same is true of a clean result, which only means nothing in that particular comparison set matched.
How do I check a document for both AI writing and plagiarism before submitting it?
Start by preparing a clean, final-format file. Turnitin's AI writing report applies to long-form English prose and requires a minimum document length, so a short abstract or a bullet-point outline will not produce a meaningful AI result [3]. Export the same document you intend to submit — headings, references and all — because the report separates flagged from not-flagged text within the actual document structure, and stripping out your bibliography changes the picture you are reviewing [3].
Then read the AI report as a breakdown, not a headline. The report gives an overall percentage alongside a section-by-section split of flagged versus not-flagged text, which tells you where the signal sits rather than just how large it is [3]. Pay particular attention to a *% style display: that appears when detection falls below Turnitin's confidence threshold, and it signals a low-confidence result rather than a clean pass [3]. A document showing *% is not certified as human-written; it is simply one the detector could not call with confidence.
Next, open the similarity report and work through the match list source by source. Read each highlighted span in context and ask a single question: is this quoted and cited correctly? Properly attributed quotation is not misconduct, and a high index built from correct citations is a formatting observation, not an integrity problem [2]. What you are hunting for is unattributed overlap — paraphrased passages that follow a source's structure too closely, or blocks of text lifted without quotation marks.
Finally, compare the two reports against each other. A pattern where the same sections are both heavily matched and heavily AI-flagged is the strongest signal that those passages need rewriting rather than patching. Because the AI report students preview uses the same format instructors later review in the LMS, a pre-submission check gives you a genuine rehearsal of the feedback conversation rather than a different metric entirely [3]. Fix the passages, then re-check — one pass rarely resolves everything.
What do Turnitin's AI score and similarity index actually tell you?
The similarity index tells you what proportion of your document's text matches sources in the comparison database [4]. That is a statement about overlap, not about intent, and it cuts both ways. A high index can be entirely legitimate — quotations, common phrasing and reference lists all generate matches — while a low index does not guarantee originality, because any source outside the database is invisible to the comparison [4]. This is why the index should always be read together with the match list and the AI writing indicator rather than in isolation [4].
The AI score tells you something narrower still: the proportion of qualifying prose the detector classified as likely machine-generated [1]. It is not a measure of how much help you received, how much you edited, or whether your work is your own. Turnitin is explicit that the indicator is a signal for human review, and that reviewers should weigh it against drafting history, version control and the assignment context rather than acting on the number alone [1]. A student who drafted by hand but ran the text through a grammar tool can still trip a flag; a student who used a model and then rewrote substantively may not.
Both numbers are also threshold-sensitive in ways students routinely misread. The AI indicator's confidence floor means low-signal results surface as *% rather than a precise figure [3], and the similarity index shifts with the size of the comparison database and the nature of the assignment — a literature review will always match more than a lab report. Neither figure is portable across institutions, disciplines or submission dates.
The most useful reading is comparative and diagnostic. Treat the two reports as a map of where your writing is exposed: matched spans tell you which citations to tighten, AI-flagged spans tell you which sentences to rewrite in your own voice with your own reasoning. Used that way, the numbers stop being a verdict you are waiting for and become a revision checklist you control [1][4].
Reading both reports side by side is far more useful than guessing — but you can only do that if you can actually see them before the deadline. That is exactly what turnitin0 gives you: a pre-submission check that returns the same two PDFs your instructor will open, an AI detection report and a similarity report, generated from your own draft in under 15 minutes in 98% of cases. Files are checked without being added to Turnitin's student paper database, so your draft stays yours, and you can delete it from your account afterwards. With 100,000+ reports delivered to 20,000+ students across the US, UK, Canada, Australia, New Zealand and Ireland, it is the fastest way to see your own two numbers before they matter.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Does one check cover both AI writing and plagiarism?
No. Similarity checking compares your text against source databases, while AI detection analyses the style of your prose; Turnitin reports them as separate layers and does not let one replace the other [1]. You need both reports to see your full exposure.
Is a high similarity score automatically plagiarism?
No. Quotations, common phrasing and reference lists all produce matches, so each highlighted span must be inspected for correct quotation and citation before any conclusion is drawn [2]. A high index is a prompt to review, not a finding.
What does a *% AI score mean?
It means detection fell below Turnitin's confidence threshold, so the result is a low-confidence signal rather than a clean pass [3]. Treat it as "the detector could not call this" and review the flagged breakdown yourself.
Can I check my draft without it being stored in Turnitin's database?
Yes, with a non-repository pre-submission check such as turnitin0, which checks your file without adding it to Turnitin's student paper database and lets you delete it from your account. Your institution's own submission will still go through its normal repository settings.
Why does my AI score change between two checks?
The indicator is threshold-sensitive and depends on the exact text submitted, so edits, reformatting and added sections shift the figure [3][4]. Re-check after substantive rewrites rather than comparing scores across different drafts.