Why is Turnitin flagging my references

In current Turnitin reports, reference lists and bibliographies are not supposed to receive AI writing flags at all. The AI writing detection model excludes bibliographies from its analysis. If your references show highlights on a report you are looking at right now, there are three realistic explanations, and none of them is "Turnitin detected AI in your reference list." The three are: the report is stale (pre-August 2023), the flags are similarity matches rather than AI flags, or the flagged text is actually prose adjacent to your references and is being analyzed as prose by the model.

We will walk through each. But the place to start is the change Turnitin made and the sentence they asked you to act on.

For the first several months after Turnitin launched AI writing detection in April 2023, bibliographies were not excluded from the analysis. A bug occasionally highlighted AI writing inside references listed in a bibliography. We know this because Turnitin documented it in their AI writing detection model release notes:

"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."

The next sentence on the same line of the release notes is:

"Resubmit to reprocess existing submissions that contain highlighted reference sections."

That second sentence is the one most students miss. The fix does not apply to already-generated reports. If your instructor generated a report before the August 2023 release and never resubmitted the paper, the old report is the one sitting in the inbox with highlights on your bibliography. The report does not auto-update.

The same release notes also contain a related change about tables. Before August 2023, longform prose inside tables was not processed. Turnitin's release notes say: "We are now able to process long-form prose text in tables." And again, the next sentence is: "Resubmit to reprocess existing submissions that contain tables."

If you have a paper with tables in it and your report predates August 2023, the same advice applies. The fix shipped, but it only takes effect for reports generated after the fix.

There is a second confusion that produces "Turnitin flagged my references" complaints. It is the most common one we see.

Turnitin produces two different reports. One is the AI Writing Report. The other is the Similarity Report. They are separate. The Similarity Report shows where your text matches text in other sources. The AI Writing Report shows where the model thinks text is AI-generated. From Turnitin's AI writing detection capabilities FAQ:

"The Similarity score and the AI writing detection percentage are completely independent and do not influence each other."

And from the same FAQ, one line further down:

"AI writing highlights are not visible in the Similarity Report."

The reverse is also true. Similarity highlights are not visible in the AI Writing Report.

Reference lists are one of the most common things to produce similarity matches. You cited published papers. Other students cited the same published papers. The titles, the author names, the journal names, and the DOIs are all identical strings. When your references show up with highlighted text, the first thing to check is which report you are looking at. If it is the Similarity Report, the highlights are not AI flags. They are matching text. The similar source could be another student paper, a published article, or a website. None of that has anything to do with AI writing detection.

There is a third mechanism worth understanding. Turnitin's AI model does not analyze every word in your submission. It analyzes what the FAQ calls "qualifying text":

"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."

The next sentence on that page is:

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

The practical meaning for a reference section is that a list of bibliography entries, even if formatted as prose, may not be treated as qualifying text by the model at all. The entries are short. Many are not full sentences. The model is built to read prose. A formatted reference list is closer to a structured list than to a paragraph.

This is also why the percentage you see on the AI Writing Report and the amount of highlighted text can look like they do not add up. A document with a long bibliography, a list of figures, and a table of contents has a large portion of non-qualifying text. The percentage is computed over the prose that the model does read. The highlights are placed on prose segments that scored above the model's threshold. If your bibliography has 40 entries and only two paragraphs of your own text get flagged, the "2%" on the report is 2% of qualifying prose, not 2% of the full document.

If you are looking at a report right now and your references appear to be flagged, here is what to check, in order.

First, check the date your report was generated. If the report predates August 2023, it was produced before the bibliography fix shipped. Nothing is wrong with the current model. The report is stale. The fix is to ask your instructor to resubmit the paper. Quoting the release note is often enough. The note is on Turnitin's own documentation page, and it says "resubmit to reprocess existing submissions that contain highlighted reference sections."

Second, check which report you are looking at. If the highlights are in the Similarity Report and not the AI Writing Report, those are similarity matches, not AI flags. Your reference list matching another paper's reference list is expected behavior for similarity checking. It does not mean anything about AI authorship. AI writing highlights "are not visible in the Similarity Report," and similarity highlights are not visible in the AI Writing Report. The two are separate.

Third, check the text immediately around your references. If the flagged text is a sentence or two of prose right before or after your bibliography (an acknowledgment, a methodological note, an introduction to the references section), the model may be reading it as prose. That is what the model is supposed to do. The flag is not on your references. It is on the prose near them. Read the highlighted text itself, not the section it falls in.

For completeness: Turnitin also observed in 2023 that false positives were more common in the first few and last few sentences of a document. This was the kind of introduction or conclusion content that reads generically. From the May 24, 2023 release notes:

"Since launch, we have observed a higher incidence of false positive detection in the first few or last few sentences of a document. Many times these sentences consist of introduction or conclusion content written in a generic way. As a result, we have changed our detection logic to help reduce these false positives."

This is relevant because students sometimes find older advice online saying "Turnitin flags introductions and conclusions." That was a known issue, and it was a 2023 improvement, not a current one. If you see that advice in a forum post from 2023, the fix has already shipped. The release note says so in its own next sentence. For current reports, this is not something to worry about as an open problem.

The summary for anyone looking at a flagged reference list right now:

1. Check the report generation date. Pre-August 2023 reports need to be resubmitted.
2. Check whether you are looking at the AI Writing Report or the Similarity Report. Similarity flags on references are expected. AI flags on bibliographies are not.
3. If the report is current and the flags are in the AI Writing Report, read the actual highlighted text. It is probably prose near the references, not the references themselves.

The August 2023 fix is documented on the page we linked. If an instructor needs to resubmit, the release notes contain the instruction. "Resubmit to reprocess existing submissions that contain highlighted reference sections."