Why Creative Writing Gets Flagged as AI by Turnitin

Creative writing can trigger false positives on Turnitin's AI detector for several reasons documented by Turnitin itself. The model analyzes what it calls "qualifying text," which includes only prose sentences written in standard grammatical sentences. 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." Short creative pieces face an additional problem: the FAQ says that in documents of 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." The FAQ also lists characteristics that produce false positives, including "content without a lot of structural variation" and "text that literally repeats itself." Creative writing that uses repetitive motifs or uniform sentence structures can match this profile even when fully human-written.

The FAQ defines what the model examines:

"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."

One Turnitin help article provides a broader list of text types the model does not reliably detect:

"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."

This matters for creative writing assignments. If your piece includes poetry fragments, script-style dialogue, or unconventional formatting, those sections may not be analyzed at all. The score you receive reflects only the qualifying prose portions, which can create a gap between what you wrote and what the percentage represents.

Creative writing assignments are often short. A 500-word flash fiction piece or a 300-word vignette falls into a range where the model struggles. The FAQ says:

"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."

The next sentence: "This means that some text that is a mix of AI-generated and original content could be flagged as entirely AI-generated."

A release note from December 2023 adds: "Results show that our accuracy increases with a little more text so submissions that contain less than 300 words may result in an AI writing score that is likely less accurate."

If your creative piece is under 300 words, the score is likely less accurate. If it is under a few hundred words, the prediction can swing to an extreme, flagging mixed or fully human-written text as entirely AI-generated. Why Turnitin says your work is 100% AI traces that swing segment by segment.

The FAQ lists text characteristics that tend to produce 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."

Several of these characteristics overlap with common creative writing techniques. A piece that uses deliberate repetition as a stylistic device can match "text that literally repeats itself." A piece with a consistent narrative voice and uniform sentence rhythm can match "content without a lot of structural variation." The official advice is to take the percentage "into consideration" rather than treating it as definitive. Single stylistic habits get read the same way, which is the question behind is the em dash an AI tell.

The detection model processes your text by splitting it into overlapping segments and assigning each a probability score:

"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."

In a longer document, segments overlap, which means a single sentence can receive multiple scores that are pooled together. This pooling smooths out individual prediction errors. In a short creative piece, there may be only one segment with no overlap, which is why the prediction becomes all-or-nothing. What AI detectors measure beyond perplexity and burstiness covers what each segment is scored on.

To summarize what we have covered:

• Turnitin's model only analyzes qualifying prose sentences. Poetry, scripts, and non-prose formatting may not be analyzed.
• One Turnitin help article lists poetry, scripts, and annotated bibliographies as text types the model does not reliably detect.
• Short documents of a few hundred words get all-or-nothing predictions with no overlapping segments to smooth errors.
• Documents under 300 words are likely less accurate, according to a December 2023 release note.
• The FAQ advises instructors to take the percentage "into consideration" when text has little structural variation or repeats itself.
• Creative writing techniques like deliberate repetition and consistent voice can match false-positive-prone characteristics.

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