Key points
- The score estimates percentage of AI-generated text at the document level, not sentence level in the base product.
- Turnitin has published its own limitations, including reduced reliability on heavily-edited or mixed human/AI text.
- Institutions vary in how much weight they place on the score in academic integrity proceedings.
Turnitin's own documentation describes the indicator as an estimate, not a determination, and recommends instructors use it as one data point among several.
What this means for you
The score estimates percentage of AI-generated text at the document level, not sentence level in the base product. Understanding this distinction is the foundation for evaluating any specific tool's marketing claims — see our full humanizer comparison table for how individual tools stack up.
advertisement