Why Annotated Bibliographies Get Unreliable AI Scores in Turnitin

One Turnitin help article says the model has known limitations with certain formats:

"The model does not reliably detect AI-generated text in the form of non-prose, such as poetry, scripts, or code, nor does it detect short-form/unconventional writing such as bullet points, tables, or annotated bibliographies."

The next sentence: "This means that a document containing several different writing types would result in a disparity between the percentage and the highlights."

Annotated bibliographies are named explicitly in this list. The model does not reliably detect AI-generated text in this format. This is a coverage limitation acknowledged in the help article, not a configuration issue on your end. If your annotated bibliography receives a high AI score, the score comes from a model operating outside its reliable coverage range.

The bibliography itself is not just unreliable. It is excluded from processing entirely. A 2023 release note states:

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

This means the bibliography entries (the citations themselves) are not analyzed. If you submitted a document before this fix and saw highlights on your reference list, resubmitting would reprocess the document without the bibliography. Why is Turnitin flagging my references and citations goes through how far the exclusions actually reach. The same release note also addressed table content:

"We are now able to process long-form prose text in tables."

The next sentence: "Resubmit to reprocess existing submissions that contain tables."

For annotated bibliographies, the citation entries are excluded. The annotation paragraphs (the prose you write about each source) are the qualifying text that gets processed. The score you see reflects only the annotation prose, not the citations.

The detector processes only qualifying text. The documentation explains:

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

For annotated bibliographies, this creates a specific dynamic. The bibliography entries (formatted citations, hanging indents, DOI links) are excluded. The annotation paragraphs are prose sentences, so they qualify. The percentage on the report reflects only the annotation prose. If your annotations are 300 words out of a 1000-word document, the percentage covers those 300 words, not the full document. That gap between percentage and page is the first thing to check when reading a Turnitin AI writing report.

The same documentation also notes:

"The model does not reliably detect AI-generated text in the form of non-prose, or code, nor does it detect short-form/unconventional writing such as bullet points (short non-sentence structures)."

The next sentence: "This means that a document containing several different writing types would result in a disparity between the percentage and the highlights."

An annotated bibliography is exactly this kind of mixed document. The disparity between the percentage and the highlights is a known behavior.

The detection pipeline segments qualifying text into overlapping sections:

"When a paper is submitted to Turnitin, sentences from the submission are extracted and segmented into overlapping sections for prediction analysis. Each segment is classified by the AI detection model and given a value between 0 and 1, denoting the probability of the text being likely human or AI-generated. Each qualifying sentence within these segments inherits the segment's score. Since segments overlap, some sentences may have multiple scores, which are then pooled into a single score. These sentence scores are further aggregated and used to compute the overall document AI writing score."

For annotated bibliographies, only the annotation prose sentences go through this pipeline. The citation entries never enter it. The segment scores reflect the probability assessment of the annotation text alone. The overall percentage is an aggregation of those annotation-only sentence scores.

Annotation prose tends to follow a predictable formula: you summarize the source, evaluate its credibility, and describe its relevance. This repetition matches patterns the documentation associates with false positives:

"Sometimes false positives (incorrectly flagging human-written text as AI-generated), can include content without a lot of structural variation, text that literally repeats itself, or text that has been paraphrased without developing new ideas."

The next sentence: "If our indicator shows a higher amount of AI writing in such text, we advise you to take that into consideration when looking at the percentage indicated."

Annotations often have low structural variation because the summary-evaluation-relevance formula repeats across entries. The phrasing repeats because students use similar verbs ("argues," "claims," "demonstrates") across annotations. The ideas are paraphrased from the source without developing new arguments, and paraphrasing on its own tends to push the number the wrong way, as why paraphrasing makes your Turnitin AI score go up sets out. Turnitin's own guidance says to take the percentage into consideration rather than accepting it at face value when the text matches these patterns.

To summarize what we have covered:

• One Turnitin help article says the model does not reliably detect AI-generated text in annotated bibliographies.
• Bibliographies are excluded from processing. Only annotation prose is analyzed.
• The percentage reflects only the annotation prose, not the full submission.
• Annotation prose tends to match false-positive patterns: low structural variation, repetitive phrasing, and paraphrased ideas.
• Turnitin advises: "If our indicator shows a higher amount of AI writing in such text, we advise you to take that into consideration when looking at the percentage indicated."
• If you had an older submission with highlighted references, resubmitting will reprocess the document without the bibliography.

If you receive a Turnitin AI report on an annotated bibliography and want to address the flagged passages, import the report and work on them. Eligible passages can be re-run at no charge.