Direct answer
Yes, an online AI detector can scan a PDF and return an AI-writing score, but the number it gives you is an indicator rather than a verdict. Turnitin's AI writing detection processes submitted documents and reports an overall percentage of AI-generated text alongside sentence-level highlights, so a PDF draft can be checked before it reaches an instructor [1]. What matters is knowing which tool produced the score, how that tool handles low-confidence results, and what the report actually shows you.
Can an AI detector check an online PDF and give a reliable AI score?
A PDF is a container, not a plain-text file, so the first thing an AI detector has to do is extract readable text from it before any analysis begins. Turnitin's detection feature operates on files submitted through its workflow and surfaces an overall AI-writing percentage together with highlighted passages, which is why a converted PDF can still produce a usable report as long as the text layer is intact [2]. Scanned image-only PDFs are a different case: if the characters are pictures rather than text, there is nothing for the model to analyse and the result will be empty or unreliable.
Reliability is the part students most often misread. The AI writing report is designed to show the proportion of qualifying text flagged as AI-generated, not to certify authorship, and Turnitin explicitly frames the score as a signal that requires human interpretation rather than proof of misconduct [2]. A high score on a PDF can reflect heavily templated prose, a literature-review section written in a formulaic register, or a document that was drafted with an assistant and then edited. A low score does not automatically mean the writing is safe either, because detectors work on statistical patterns rather than on knowledge of who typed the words.
The practical consequence is that you should treat any online PDF check as a preview of how a reader using the same detector will see your document. If you are checking outside your institution, confirm which engine is behind the tool — a generic "AI detector" that uses an unrelated model can disagree sharply with the report your tutor actually opens [2]. Matching the detector to the one your university uses is the difference between a meaningful rehearsal and a misleading number.
Why does a Turnitin AI report show an asterisk (*%) instead of an exact percentage?
The asterisk is a confidence marker, not a formatting glitch. When the share of detected AI writing falls below Turnitin's 20% confidence threshold, the report replaces the exact figure with an asterisk so that readers do not over-interpret a number the model cannot stand behind [3]. In other words, a *% result means the detector found some signal but not enough of it to report a precise percentage with confidence.
This design choice matters enormously for anyone checking a PDF online. Students often see *% and assume the document is either completely clean or that the tool has failed, when the honest reading is that the AI signal is weak and low-confidence [3]. The same logic applies at the other end of the scale: a report showing very little flagged text is not a guarantee of human authorship, because detection is probabilistic and thresholds exist precisely to stop borderline results from being presented as certainties.
For your own workflow, the asterisk tells you where to spend your effort. If a section of your PDF returns *%, the sensible response is to review that passage for the register a detector reacts to — uniform sentence rhythm, generic transitions, list-like structure — rather than to chase the number down to zero [3]. If a section returns a firm high percentage, that is the passage worth rewriting or evidencing with your own analysis, quotations, and citations.
How can I preview my own Turnitin AI and similarity report on a PDF before I submit?
Self-checking a draft is a legitimate study habit, and the mechanics depend entirely on what your institution has enabled. Where a self-check submission is available, you upload the file, receive a similarity report and — where the feature is switched on — an AI writing report, and you can revise before the graded submission [4]. The important caveat is repository behaviour: depending on how the assignment is configured, a self-checked file may be stored in the student paper database, which can affect a later similarity match [4].
That caveat is the reason many students look for an independent pre-submission check instead of using the institutional route. A non-repository check lets you see the same two artefacts an instructor sees — the similarity report with matched sources and the AI writing report with flagged passages — without adding your draft to a database that later submissions are compared against [4]. It also lets you check as many times as you need while you revise, which the institutional self-check usually does not.
The most useful sequence is to export your final draft to PDF, run the check, and read the report the way a marker would: start with the similarity summary, then open the AI report and look at which paragraphs are highlighted and why. Fix the citation gaps first, since those are objective and fixable, then address flagged passages by adding your own argument, data, or paraphrase in your own voice [4]. Re-checking after revision is what turns a single score into an actual improvement loop.
If you would rather see the report before your tutor does, turnitin0 runs a non-repository check on your PDF and returns both the AI writing report and the similarity report as downloadable PDFs, usually in under 15 minutes — the same two documents your instructor opens in the LMS, with the score, the flagged passages, and the matched sources visible on your own draft.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
Get Real [Turnitin](https://www.turnitin0.com/) AI & Similarity Report
FAQ
Does a PDF lose text when an AI detector reads it?
Only if the PDF has no text layer. A PDF exported from Word, Google Docs, or LaTeX keeps its characters and is analysed normally, whereas a scanned image PDF must be OCR'd first or the detector will find nothing to score [1].
Is a low AI score on a PDF proof that the writing is human?
No. Detection is statistical and threshold-based, so a low or asterisk result means the signal was weak, not that authorship has been verified [3]. Keep your drafts, notes, and version history as the real evidence of your process.
Can my university see that I checked my draft online?
An independent check is not connected to your institutional account, and a non-repository check does not add your file to the student paper database [4]. Institutional self-checks, by contrast, may store the file depending on assignment settings.
What should I do if only one paragraph is flagged?
Treat it as a revision target rather than a crisis. Rewrite that paragraph with your own analysis, add specific evidence or citations, and vary the sentence structure, then re-check to confirm the change registered [2].
Does the AI score affect the similarity score?
They are separate reports. Similarity measures text matching against sources and the paper database, while the AI report measures statistical patterns associated with generated text, so one can be high while the other is low [2].