How do I interpret results from GPTZero's Advanced Sentence Scanning?
Last updated: October 21, 2026
Advanced Scan shows which sentences had the strongest effect on the result. Those sentences are highlighted in the document.
Yellow means the sentence pushed the result toward AI. A darker yellow is a stronger effect.
Green means the sentence pushed the result toward human. A darker green is a stronger effect.
The same sentences appear under Sentences Impacting the Result, arranged from High AI Impact on the left to High Human Impact on the right.
Some sentences are not highlighted. They can still affect the result, but not more than the rest of the document. Editing the whole document can move the score more than editing only the highlighted sentences, because the model reads the document as a whole.
Understanding the result
The result has a class and a confidence. The class is AI, Mixed, or Human. A highly confident AI result reads: “We are highly confident this text was AI-generated.”
The percentage is the model’s confidence in that classification of the entire text. “100% AI” means the model is confident the whole document was written by AI. It does not mean that 100% of the sentences were written by AI. Mixed and Human use the same rule.
A document written by a person gets the same card with a Human class. Here, the model is highly confident the text is entirely human, and the Human score is 100%.
A document that combines AI and human writing gets a Mixed class. The card also tells you how many sentences are likely AI-generated.
See How do I interpret a GPTZero result? for how to read the score.
Unlike the plagiarism check, which looks for matching text, AI detection is a prediction. AI writing and human writing both keep changing.
That is one reason not to use an AI detection result as the only basis for academic punishment or discipline.
Read more on what to do if you detect AI in your classroom.
Understanding confidence
Every score also comes with a statement about our confidence in the prediction. The wording is highly confident, moderately confident, or uncertain. Confidence measures the rate at which we would incorrectly identify the text document (error rate). These can range from:
Highly confident: our error rate is less than 2%
Moderately confident: our error rate would be around 10%
Uncertain: our error rate would be 14% or higher
The wording on the card changes with the confidence level. Highly confident results look like the AI and Human examples above. A moderately confident result reads “We are moderately confident this text was AI-generated”, and the scores are split between classes.
An uncertain result reads “We are uncertain about this document. If we had to classify it, it would likely be considered human.” The scores are close to even. Treat it as no clear signal either way.
Our confidence levels adapt automatically to maintain the highest accuracy rate (95%+) while keeping the lowest false positive ratio (1%). In other words, if we aren’t confident something is AI, we generally err on the side of “human” with a low confidence score to avoid falsely accusing a human of being AI.