A high-level meaningful feature (like realistic gaze patterns) used to detect or classify content, as opposed to low-level pixel artifacts.