Does Turnitin detect AI in a PDF?

Yes. PDF is one of four file types on Turnitin's accepted list for an AI Writing Report, alongside .docx, .txt and .rtf. Submitting as a PDF does not remove your document from AI writing analysis. What the format does change is what your file contains, and that has real consequences elsewhere in your submission.

Straight from the file requirements for an AI Writing Report:

"Accepted file types: .docx, .pdf, .txt, .rtf"

The same short list also sets the language rule: "File must be written in a supported language: English, Spanish, Japanese". If that part is your question rather than the format, we took it apart separately in can Turnitin detect AI in non-English submissions.

That is the entire format answer. There is no separate PDF pathway, no lighter treatment, nothing in the published requirements that treats one accepted type differently from another.

The same thing it reads inside a Word file: prose. Turnitin calls it qualifying text, and describes it on the page about using the AI Writing Report.

"This qualifying text includes only prose sentences, meaning that we only analyze blocks of text that are written in standard grammatical sentences and do not include other types of writing such as lists, bullet points (short non-sentence structures), or other non-sentence structures."

Then: "This percentage is not necessarily the percentage of the entire submission."

So the percentage on a thesis PDF is a statistic over the running prose inside it, not over the pages. Turnitin says elsewhere that a document containing several writing types "would result in a disparity between the percentage and the highlights", which is the documented reason your highlighted passages never look like they add up to the number. More on reading that mismatch in how to read a Turnitin AI writing report.

Two specifics that matter for anything thesis-shaped, both from Turnitin's AI writing detection model release notes.

Prose inside tables is processed: "We are now able to process long-form prose text in tables." The following line adds the condition that gets forgotten: "Resubmit to reprocess existing submissions that contain tables." Existing reports do not update themselves.

Bibliographies are excluded: "We have fixed a bug that was occasionally highlighting AI writing within references listed in a bibliography. Bibliographies are now excluded when processing the AI writing report." Same catch, same wording: "Resubmit to reprocess existing submissions that contain highlighted reference sections."

If you are looking at a report with a highlighted reference list, check the date it was generated before you conclude anything about your references. Also check which report you are holding: Turnitin's documentation says AI writing highlights "are not visible in the Similarity Report" at all, and matching reference entries are the ordinary case in a similarity report, since a citation style exists to make everyone format the same source the same way.

Not the detection. The document.

Exporting to PDF flattens the machinery in a Word file: fields, the generated table of contents, cross-references and automatic numbering stop being live objects and become printed text. That matters when you still have revisions ahead of you, and it matters more when the round trip goes the other way, because bringing text back out of a PDF and into a document is where numbering and structure get quietly broken (what breaks in a long document).

The practical rule is boring. Keep revising in the source file. Export the PDF last, from the version you are actually submitting.

One line worth carrying:

"In shorter documents where there are only a few hundred words, the prediction will be mostly 'all or nothing' because we're predicting on a single segment without the opportunity to overlap."

And its follow-up: "This means that some text that is a mix of AI-generated and original content could be flagged as entirely AI-generated." A one-page PDF is not a smaller version of the same measurement. It is a blunter one.

Two separate gates, and neither is about your file. Turnitin says AI writing detection "is only be available to customers that license Turnitin Originality" (their grammar), with an add-on route for iThenticate 2.0 accounts. And it says only instructors and administrators can see the indicator, adding that "with the PDF download feature, instructors can download and share the AI report with students".

That last part is the PDF that most students actually end up holding: a PDF of the report, not a PDF they submitted. Worth being precise about which one is in the conversation.

Is submitting a PDF safer than submitting a Word file? No. Both are on the same accepted list, and the analysis described in Turnitin's documentation does not distinguish between them.

What about a PDF that is only scanned page images? There is no prose to extract from an image, and treating that as a way through is a bad plan on its own terms. A page-image PDF cannot be searched, quoted or annotated by whoever is reading it, which is usually a submission requirement somewhere in the paperwork. And Turnitin's own line is that the AI score "should not be used as the sole basis for adverse actions against a student", which means the person, not the statistic, is doing the deciding. Removing the statistic does not remove the reader.

My PDF got a similarity score but no AI percentage. The two are independent. Turnitin's documentation says the similarity score and the AI writing percentage "are completely independent and do not influence each other". Missing AI numbers point at the licence, the language rules or who is allowed to see the indicator, not at your file (the difference in full).

Does a PowerPoint export to PDF get an AI report? The PDF is on the accepted list, so mechanically yes. But slides are mostly bullets, and bullets are not qualifying text, so the prose the report has to work with may be a small fraction of what is on the slides.