Why Did Your Turnitin AI Score Change on Resubmission?

Turnitin's AI writing score is not a permanent property of your document. It is the output of a detection model applied at the moment of submission. When the model changes, the score can change too. According to the documentation, "As we iterate and develop our model further to better detect newer LLMs, it is likely that our detection capabilities will also change, affecting the AI percentage.. However, for a submitted document, the AI percentage will change only if it's re-submitted again to be processed."

So if you resubmit and the number moves, that is expected behavior. The document did not change. The model did.

One common source of confusion is the assumption that an old report will refresh itself when Turnitin updates its model. It will not. The documentation is explicit: "This release will not retroactively update previously-generated AI writing reports. To view the AI writing score generated by this update, existing submissions must be resubmitted."

This means a report generated three months ago still shows the score from the model that was active at that time. If you want to see what the current model says, you have to submit the document again. There is no silent update in the background.

This matters for anyone comparing scores across time. If a student submitted in March and resubmitted in August, the two numbers are not directly comparable. They were produced by different versions of the detection model, which is why comparing two reports rather than two scores is the only reading that holds up.

The FAQ says that in July 2026, Turnitin updated its model architecture to consolidate a multi-model ensemble into a single model. The FAQ says this update improves and simplifies the AI writing report while maintaining a false positive rate under 1%, a figure whose real-world size is worked out in what a 1% false positive rate means.

This consolidation is a structural change to how detection works. A multi-model ensemble means multiple detectors each produce a score, and those scores are combined. Moving to a single model means one detector now produces the score directly. The FAQ says the change was made to improve and simplify the report, but a side effect is that resubmitted documents can produce noticeably different percentages than they did under the old ensemble.

We mention this not to alarm anyone, but to set expectations. If your score changed after July 2026, the model consolidation is a likely explanation.

To understand why a score can shift, it helps to know how the percentage is calculated. 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.

The final percentage on your report is the result of many small, sentence-level predictions aggregated into a single number. When the model updates, the predictions on individual segments can shift. A sentence that scored 0.4 under the old model might score 0.6 under the new one. None of these shifts are large on their own, but when aggregated across an entire document, they can move the overall percentage by several points.

This is why two submissions of the same document can produce different scores. The text is identical, but the model evaluating each segment has changed.

If you resubmit a document and the AI percentage goes up, that does not necessarily mean the text got "more AI." The same surprise after a round of edits is covered in I edited my paper and the AI score went up. It means the updated model is more sensitive to patterns in those segments. If the percentage goes down, the updated model may be less sensitive to those same patterns. Either way, the change reflects a difference in the detector, not a difference in the writing.

Here are the practical takeaways:

• A score from a previous submission is frozen in time. It will not change unless you resubmit.
• A new score reflects the current model, which may be different from the one that produced the original report.
• Comparing scores across different model versions is not an apples-to-apples comparison.
• If you are revising a document to lower its AI percentage, always work from the most recent submission, since that is the score the current model will produce.

If your AI percentage shifted after resubmission, focus on the flagged passages rather than the number itself. The report identifies specific sentences and segments that the model classified as AI-generated. Those are the sections to revise, in the order rewriting only the paragraphs a Turnitin report flagged sets out.

Keep your citations, references, and formatting in place. Only the flagged passages need attention. After revising, submit again to get a fresh score from the current model.

Eligible passages can be re-run at no charge, so you can refine flagged sections without additional cost until the percentage reaches an acceptable level.