What Turnitin AI Detection Checks: How the Score Is Built

Turnitin's AI writing detection is a separate analysis from the similarity (plagiarism) check. The similarity score looks for text that matches existing sources. The AI writing score looks for statistical patterns in word choice that resemble machine-generated text. Turnitin's FAQ says the two are "completely independent and do not influence each other." The AI detector works by splitting your submission into overlapping text segments, classifying each segment by the probability that it was human or AI-generated, and aggregating those scores into a document-level percentage. Only prose sentences are analyzed. Lists, bullet points, and other non-sentence structures are excluded. And only instructors and administrators can see the AI score, unless the instructor chooses to share the PDF report.

Turnitin's FAQ describes the pipeline this way:

"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."

Read this as a pipeline. Your text goes in, gets cut into overlapping pieces, each piece gets a probability score, sentences inherit scores from their segments, overlapping scores get pooled, and the pooled scores get aggregated into one number. The model is not reading your argument or checking whether your ideas are original. It is reading statistical features of your word sequences.

Two statements from the same FAQ clarify what the model does and does not measure. First: "Our model is not explicitly programmed to evaluate specific signals such as 'burstiness,' 'perplexity,' or other individual metrics sometimes referenced in public discussions." The next sentence reads: "Instead, it learns statistical patterns from our training data." Second: "Our classifiers are trained to detect these differences in word probability and are adept at the particular word probability sequences of human writers."

The first statement says the model does not compute two named metrics. The second says the model is trained on word probability. These are consistent. The model does not calculate a number called perplexity and compare it to a threshold. But it does learn from word probability data. The distinction matters because a lot of advice about avoiding AI detection is built on the assumption that you just need to move a named metric. If the model does not compute that metric, adjusting it is not the lever people think it is. What AI detectors measure beyond perplexity and burstiness takes that further.

Only prose sentences count. The FAQ states:

"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."

This means that if your document contains a lot of lists, bullet points, or other non-prose content, the AI percentage applies only to the prose portion, not to the whole document. The percentage you see may represent a smaller portion of your text than you assume.

Scores between 1% and 19% are not displayed as numbers. The FAQ explains:

"To avoid potential incidence of false positives, no score or highlights are attributed for AI detection scores in the 1% to 19% range. When AI is detected below the 20% threshold in the report, it is now indicated with an asterisk (*%) and no percentage is attributed."

If your report shows an asterisk instead of a number, the detector found some AI-like patterns but not enough to warrant a specific percentage. This was a deliberate design choice to reduce false positives in the low range. What the asterisk (*%) means on a Turnitin AI score has the rest of it.

The FAQ is explicit about access: "only instructors and administrators are able to see the indicator." Students cannot view their own AI score directly. The FAQ continues: "However, with the PDF download feature, instructors can download and share the AI report with students." If you are a student and want to know your score, you need to ask your instructor to share the PDF version of the report — and then how to read a Turnitin AI writing report is the next step.

To summarize what we have covered:

• Turnitin's AI score and similarity score are completely independent measurements.
• The AI detector splits text into overlapping segments and classifies each by word probability patterns.
• Only prose sentences are analyzed. Lists, bullet points, and non-sentence structures are excluded.
• Scores between 1% and 19% show only an asterisk, not a number, to reduce false positives.
• Only instructors and administrators can see the AI score, but they can share the PDF with students.

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