Does Turnitin Read Text Inside Tables? The Table Prose Rule

If you have ever submitted a paper with data tables and then stared at the AI writing report, you have probably wondered what Turnitin actually analyzed. The confusion is understandable. For a long time, one Turnitin help article says: "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."

That wording placed tables in the same category as poetry and scripts, formats the detector would not touch. Students and instructors interpreted this to mean that any text inside a table was completely ignored. That was a reasonable reading, but it was not the full picture.

The reality is that Turnitin's AI detection only processes what it calls prose sentences. 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." Tables were not excluded because of a special rule about tables. They were excluded because most table content does not qualify as prose in the first place.

Turnitin updated its processing logic. The entry dated 9 August 2023 on its AI writing detection model page (guides.turnitin.com/hc/en-us/articles/28294949544717-AI-writing-detection-model) states: "We are now able to process long-form prose text in tables." This is a meaningful change. If your table contains full sentences written in standard grammatical form, those sentences are now analyzed by the AI detection model.

The key phrase is "long-form prose text." A table cell that reads "The results indicate a statistically significant correlation between variable X and variable Y, with a p-value below 0.01" is a standard prose sentence. That sentence will be processed. A table cell that reads "p < 0.01" or "Variable X, 0.85" is not a sentence. It is a data fragment, and it will be skipped.

This distinction matters because many academic tables mix full sentences with short data entries. The sentence portions will be analyzed. The data portions will not. You cannot assume that an entire table is either included or excluded. Turnitin evaluates content on a block-by-block basis, looking for standard grammatical sentences regardless of whether they happen to sit inside a table cell or in the body of the document.

That same entry also instructs: "Resubmit to reprocess existing submissions that contain tables." If you submitted a paper before this update, the table prose in that submission was not processed. Resubmitting triggers the updated logic.

Even with the table prose update, several types of content remain outside the scope of AI detection analysis. The prose-only rule means that bullet points, short non-sentence structures, and lists are not processed. If your table contains short fragments rather than full sentences, those fragments are excluded.

Bibliographies were fixed on the same date. The 9 August 2023 entry 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." This fix addressed a specific problem where reference entries were being flagged, which produced misleading results. The guidance is clear: "Resubmit to reprocess existing submissions that contain highlighted reference sections."

So the excluded list includes short table fragments, bullet points, data entries, and now bibliographies in their entirety. The detector focuses on prose sentences in the body of the document and in table cells, and it skips everything else. This is not a flaw in the system. It is a design choice built around the fact that the detection model was trained on prose, not on data formats or citation structures.

One of the most common sources of confusion is the gap between the AI writing percentage and the highlighted text in the report. The documentation addresses this directly: "This means that a document containing several different writing types would result in a disparity between the percentage and the highlights."

If your document includes tables with full sentences, short data fragments, bullet points, and a bibliography, only the prose sentences are analyzed. The percentage reflects the proportion of analyzed prose that the model classified as AI-generated. It does not reflect the proportion of the entire submission. The documentation notes: "This percentage is not necessarily the percentage of the entire submission."

This is why you might see a 40 percent AI writing score but only a few highlighted paragraphs. The highlights show which prose segments were flagged. The percentage is calculated from the analyzed prose only. Content that was excluded, such as table data or bibliography entries, does not appear in the highlights and does not factor into the percentage calculation in the same way you might expect.

Understanding this distinction helps you read the report correctly. The percentage is not a simple ratio of flagged text to total text. It is a score based on the prose the model actually processed, which may be a subset of your full document.

Before you submit a document that contains tables, we recommend reviewing the content of those tables. If your table cells contain full sentences, those sentences will be analyzed by the AI detector. Make sure those sentences reflect your own writing. If you used AI to draft any table prose, that prose will be evaluated just like any other sentence in your paper.

Check your bibliography. Bibliographies are now excluded from AI writing analysis, so references should not appear as highlighted text. If you have an older report where references were flagged, resubmitting the document will apply the updated logic that excludes bibliographies.

If you have a document with mixed content types, do not be surprised if the percentage and the highlighted sections seem inconsistent. As the documentation explains, disparity between the percentage and the highlights is expected when a document contains several different writing types. Read the highlights to understand which specific prose segments were flagged, and treat the percentage as a score derived from the analyzed prose rather than from the entire submission.

If you want to verify how your document will look under Turnitin's AI detection before submitting it for grading, we built a tool for that. Try our Turnitin report preview to see where your prose stands.