Every sentence scored on its own, with the reason it was flagged. A likelihood, never a verdict on who wrote something.
The text is broken into sentences and each is assessed on its own.
Each goes through the provider chain — a licensed detector where configured, otherwise model assessment.
Scores are combined into an overall figure with a confidence level.
Every flag comes with its reason, so you can disagree with it.
Read this before acting on a number. Detectors over-flag formal, careful, non-idiomatic English — which describes most writing produced by people working in a second language. A high score is not evidence of anything on its own. Open the sentence view and judge it yourself.