AI Overviews are query-dependent and personalized, not static rankings
Google describes AI Overviews, in its AI features documentation, as a generated summary that can appear above traditional results for some queries, varying by query, user context, and over time. There is no published, stable “position” the way a traditional ranking position is defined. Any tracking methodology must account for this variability rather than treating a single observation as a stable rank.
Define the query set deliberately
Pick a fixed, documented list of queries relevant to the business, not queries chosen because they happened to show an AI Overview in a one-off check. Record why each query was included. A query set assembled after seeing favorable results is not a methodology, it is selection bias.
Control for session and location variables
AI Overview appearance and content can vary by signed-in state, location, device, and query phrasing. Run checks from a consistent, documented configuration (logged out or a defined account state, consistent location setting, consistent device) and record that configuration alongside each observation. Without this, two checks run under different conditions are not comparable.
Record presence, not just position
Because AI Overviews do not publish a numbered rank, record for each query: whether an AI Overview appeared at all, whether the tracked domain was cited or linked within it, and the observation date and configuration. Treat this as a presence-and-citation log, not a ranking table, since no provider documents an equivalent to traditional position-1-through-10 ranking for AI Overviews.
Set a realistic observation cadence
AI Overview content for a given query can change across sessions and over time. A single observation is a snapshot, not a trend. Repeat checks on a fixed cadence (for example weekly) using the same query set and configuration, and only draw trend conclusions from the accumulated log, not from any single check.
What this methodology cannot prove
This approach can document observed presence and citation patterns over time for a defined query set. It cannot prove why an AI Overview did or did not cite a given domain, since Google does not publish its selection criteria, and it cannot guarantee future appearance based on past observation.
See ai citation tracking methodology, google ai overviews website readiness, and ai search content experiment.
Frequently asked questions
Can I track my AI Overview ranking the way I track search position?
Not directly. AI Overviews do not have a documented numbered position; the most defensible metric is presence and citation observed under a fixed, logged configuration.
Does a tool claiming exact AI Overview rank numbers have proprietary access Google does not publish?
No tool has documented, provider-confirmed access to an internal AI Overview ranking metric; treat any such number as a vendor derived estimate, not an official measurement.