Once a submission has gone through GPTZero, the Probability Report for the submission will return with three different detection values.

The AI Classifications will either be AI Only, Human Only, or Mixed (i.e. both Human and AI was detected).

The AI Confidence percentage is a reference to the perceived probability that the submission as a whole contains AI content.

A submission with a high percentage (e.g. 95%) means there is a high probability that AI was used. A low percentage (e.g. 3%) is a high probability that the submission was not generated by AI.

In addition each sentence is individually scored and highlighted to indicate where in the document it is believed AI occured.

Why are there no sentences highlighted when the AI Confidence shows a low percent (e.g. 4%)?

This is because each sentence is scored individually, a low score for the sentence may be too low to trigger the highlighting. Where the AI Confidence and Classification is looking at the submission as a whole.

Why are all of the sentences highlighted if the AI Confidence isn't 100%?

This can occur because each individual sentence is scored separately and highlighted if AI is detected. While the overall AI Confidence could be 93% for the submission, AI was believed to be detected in every sentence.

For additional insight into how GPTZero AI Detection is scored, please see GPTZero FAQ.