AI writing score

In plain English

If a report says 40%, it is claiming that about forty percent of the qualifying text tripped its threshold. It is not saying there is a 40% chance you used AI, and it is not saying you are 40% guilty.

That distinction sounds pedantic until you are in a meeting about it, where almost everyone in the room will read the number the second way.

A worked example

A Turnitin AI writing report, read line by line. The behaviour described here is the vendor's own documented policy, linked below.

AI writing detection            *

  ── what this means ───────────────────────────────────
  Not "0%". Not "clean". Turnitin attributes no score and
  no highlights between 0% and 20%, and shows an asterisk
  instead, because its own testing found more false
  positives in that band.

The asterisk is the most misread element of the whole report. It is not a pass, not a zero, and not an error — it is the vendor declining to publish a number it does not trust at that range.

Above the 20% threshold Turnitin states it aims to keep false positives under 1%. Below it, the company suppresses the figure entirely rather than show one that would read as more certain than its evidence supports.

So a report showing an asterisk and a report showing 0% are different claims, and a report showing 35% is a claim about how much of the document was flagged — never about how likely it is that a person cheated.

Why it matters for AI detection

Because the number is almost always the only thing anyone looks at, and the units are almost always misunderstood. A score is a measurement of the text's predictability, expressed as a proportion of the document.

Turnitin itself states the indicator is not intended as the sole basis for an academic misconduct finding. That sentence is worth quoting verbatim in any meeting where a score is being treated as the finding.

The practical consequence: a score is a reason to open a conversation and look at drafts, version history and the student's other work. It is an input to a human decision, and the vendor agrees.

Commonly confused with

Similarity score

A similarity score says how much of your text matches sources in a database. An AI writing score says how much of it reads as machine-generated. Different systems, often printed on the same report, routinely conflated.

False positive rate

The score is about one document. The false positive rate is about how often the tool is wrong across many — which is the number that decides whether the score should be acted on at all.

Read next

• What Turnitin's AI detection measures
• If you have been falsely accused
• False positive rate
• N-gram
• AI slop
• Plagiarism score vs AI score
• What the similarity score counts