The distinction that matters is between the surface and the model. A buyer opens a product, and which model happens to serve them is not something they chose or can see. Measuring the surface is measuring what a buyer actually encountered.
It also decides what an answer means. An answer written from what a model memorised in training is not what a buyer sees today and cannot be influenced by anything you ships, which is why answers are required to be grounded in a live search.
Try it on your own site, free
No card, and the free tier does not expire — the first verdict takes about six weeks to arrive.