It is the listening layer of an AI visibility program. Rather than checking the engines by hand, monitoring runs your prompts on a schedule and logs every mention: which engine, which prompt, what position, what sentiment, and which sources the engine leaned on.
This catches three things you cannot see otherwise: new gaps where a competitor started getting recommended instead of you, drops where you lost a mention you used to hold, and inaccuracies where the engine describes you wrongly. Each is an action item.