Not as a position, and any AI Overviews tracker that shows you a settled position 3 inside the overview is computing something it is not telling you about. The blue links underneath are close to stable: ask two people to search the same phrase in the same city on the same day and they see broadly the same ten results. The overview above them is generated per search, and Google decides per search whether to show one at all. In Search Console the whole overview counts as a single result, not as a list of places.
What does exist, and what Verand records, is an order inside the text. Every brand named in the overview is numbered by where its first mention falls, so first named scores 1.0 and each place below it 0.15 less, and your average position is the mean across the overviews that named you. That is a real, reproducible number. It is not a Google ranking, and this page does not call it one. Your actual Google rankings are tracked daily in the same account, against the same pages.
The honest unit above that position is a rate. Every prompt in your set is searched once a day, every day, one search per prompt per assistant per scan, and the dashboard reports the share of answers that named you, the share that cited you, and how both moved. The trend is one point per scan day over a 30-day window by default.
The number is also only as good as the searches behind it. A prompt set built from the topics you actually sell into will move when your content moves. A prompt set of branded searches will sit at 100 percent and tell you nothing, which is the reason for the one-in-five ceiling on branded prompts.
Where the search is run from. The overview is the one part of this that is geographic, so Verand runs Google’s two AI surfaces from a place rather than from nowhere. For a local business with a confirmed address in the United States or Canada that means city level, rotating across every office you have listed rather than only the head one, and a search that names one of your cities is watched from that city. Everywhere else it is country level, defaulting to the United States. Reads are desktop. The chat assistants ignore all of this, because their answers are placeless.
One limit, stated plainly. Because absence is stored nowhere, your Google AI Overviews check count will sit below your prompt count, and that is on purpose: those are the searches where Google showed no overview. An answer that came back without you in it is a different thing. That is a stored row with mentioned set to false, which is why the trend shows the scan where you dropped out. A failed check is a third thing again, kept separately with the provider’s own error message so it never reads as a miss.