Direct answer
An AI detector file check means uploading a document (.docx,.pdf,.txt,.rtf, or.html) and having the tool return an AI-writing percentage plus sentence-level highlighting on that exact file. Turnitin's AI writing detection only evaluates long-form, prose-style English text, so the file you upload must contain real, extractable text — not an image or a scanned page [1]. The report you get back mirrors what an instructor sees in the LMS: an overall score, flagged passages, and (in a full report) a similarity summary [1].
What File Types And Limits Does An AI Detector Accept, And How Does It Score An Uploaded File?
Turnitin accepts the common document formats students actually submit: Microsoft Word (.doc), Word 2007+ (.docx), OpenOffice (.odt), rich text (.rtf), plain text (.txt), HTML, and PDF [2]. The practical caps matter more than the format list. A submission must stay under roughly 100 MB and under 800 pages, and the file needs at least 20 words before a similarity report will generate at all [2]. Anything outside those bounds is rejected before detection ever runs, which is the single most common reason a student thinks the detector "isn't working."
The format caveat is the one that catches people out: a PDF is only usable if it carries a real text layer. If you scanned printed pages, photographed handwritten notes, or exported slides as flat images, Turnitin extracts no words and the AI writing report cannot be produced [2]. The same logic explains why spreadsheets, code files, and image-only submissions return nothing useful — AI detection is built for continuous prose, not tabular or visual content [1].
Once a valid file clears those gates, scoring runs in two layers. The first is a document-level percentage representing the share of qualifying text the model classifies as AI-generated [1]. The second is sentence-level highlighting inside the document viewer, so you can see precisely which paragraphs triggered the flag rather than guessing from a single number [1]. That pairing — one headline figure plus located evidence — is what makes an uploaded file report actionable instead of merely alarming.
How Accurate Is An AI Detector On A File, And What Do The Flagged Percentage And Highlighted Passages Actually Mean?
Treat the percentage as a confidence-weighted signal, not a verdict. Turnitin's own reporting displays an asterisk (%) instead of an exact figure when its AI detection confidence falls below its 20% threshold, explicitly flagging those results as low-confidence rather than precise [3]. In practice this means a file scoring % is not "20% AI" — it is the detector telling you it does not have enough signal to commit to a number [3].
The highlighted passages deserve the same restraint. Sentence-level flags identify text whose statistical patterns resemble generated prose, and those patterns also appear in heavily templated human writing — formulaic literature reviews, boilerplate methodology sections, and heavily paraphrased summaries [3]. Turnitin states plainly that its AI writing score should not be used as the sole basis for adverse action against a student, precisely because false positives are possible [3]. A flagged paragraph is a prompt to look at the source and the drafting history, not proof of anything.
Accuracy also degrades in predictable ways that a single score hides. Short files give the model less evidence, mixed human-AI documents produce blended scores that describe no single paragraph well, and non-English or non-prose content falls outside the detector's design scope entirely [1][3]. If your file is a 400-word discussion post, expect a noisier reading than a 3,000-word essay — and read the number accordingly.
Can I Check My File With Turnitin Before Submitting It, Without It Going Into The Repository?
Yes — and doing so is now normal practice rather than a loophole. Because AI detection is probabilistic and instructors increasingly run reports on final submissions, checking a draft first lets you fix genuinely AI-sounding passages while there is still time to rewrite them [4]. The thing to verify before you upload anywhere is whether the checker is repository-based: a standard institutional assignment submission can add your paper to Turnitin's student paper database, which then affects later similarity matches for you and your cohort [4].
A non-repository check is the safer shape for a pre-submission preview. It runs the same detection and similarity analysis on your document but does not retain the file in the student paper database and does not share reports with third-party databases, so the preview does not contaminate future similarity results [4]. That distinction is why students who want a genuine dry run deliberately look for a non-repository option rather than re-submitting to their course shell.
The third thing to check is turnaround and deliverable format. A preview is only useful if it arrives before your deadline and returns the actual report — the AI percentage with highlighting plus the similarity report with matched sources — not a stripped-down score [4]. If you cannot download the PDF and read the flagged passages yourself, you have not really previewed what your instructor will see.
Reading a Turnitin report on your own file is far more useful than reading a bare percentage — and turnitin0 lets you do exactly that before your deadline. Upload a.docx,.pdf, or.txt (English, over 300 and under 30,000 words, under 20 MB) and you receive two downloadable PDFs in one checkout: the AI detection report and the similarity report, matching what professors see in their LMS. Checks run without adding your file to Turnitin's student paper database, most orders finish in 5–15 minutes, and turnitin0 has delivered 100,000+ reports to 20,000+ students across the US, UK, Canada, Australia, New Zealand, and Ireland. If a flagged passage needs rewriting, the AI humanizer works on ChatGPT, Claude, and Gemini text and preserves your meaning, citations, headings, and.docx formatting.
※ Turnitin0.com - Actual [Turnitin AI Report](https://www.turnitin0.com/guides/us/is-there-a-humanizer-that-works-like-quillbot-but-doesn-t-get-flagged-by-turnitin) Cover, Score, Flag And Similarity Summary
FAQ
Can I upload a PDF to an AI detector?
Yes, provided the PDF contains a real text layer rather than scanned images. Turnitin lists PDF among accepted formats, but a PDF with no extractable text will not produce an AI writing report [2].
Why does my AI detector file score show an asterisk instead of a number?
Turnitin displays *% when its AI detection confidence is below its 20% threshold, which marks the result as a low-confidence signal rather than an exact percentage [3].
Does uploading my file to a checker add it to Turnitin's student paper database?
It depends on the checker. Institutional assignment submissions can be retained in the student paper database, while non-repository checks analyze your file without storing it there or sharing reports with third-party databases [4].
Is a high AI score on my file proof that I used AI?
No. Turnitin advises that its AI writing score should not be the sole basis for adverse action, because templated human writing can trigger the same patterns the detector looks for [3].
What is the minimum file size an AI detector will score?
Similarity reports require at least 20 words, and very short files give AI detection too little text to produce a reliable figure [2][1].