Does Search Console report AI Overviews separately from AI Mode?

Google's generative AI performance report includes impressions from AI Overviews and AI Mode. Its default view covers supported generative AI features, so the headline number should not be described as an AI Overviews-only count. Record the report name, search type, date range and any applied dimension beside every exported figure. That context prevents a combined total from becoming a claim about one feature[1].

What does an AI Overview impression measure?

In the generative AI report, an impression is a time a link to your site was shown to a user in a supported generative Search feature. It is exposure of a link under Google's counting rules, not proof that the user read the page or clicked it. A click is a separate event: someone followed a Google link to the site[1][2].

Can I filter generative AI impressions to the United States?

Yes. The report has a Countries dimension that groups data by the country where the search originated. Use the United States row or filter for the US business question, and label the figure accordingly. That view does not become worldwide exposure simply because the site is based in America. Keep country and period with the number when comparing reports[1].

Why is the generative AI report missing for my site?

Google says the report is usually absent when a site has not received enough impressions in generative AI features on Search. Its help page also notes that a site excluded from those features may lack eligibility. An absent report is therefore not a measured zero for every query or every AI service. Check the property's inclusion setting and avoid inventing a visibility total[1].

Can a third-party tracker see Google's internal ranking data?

Google says third-party tools do not have access to its internal ranking data and cannot guarantee performance. Ask a tracker what it sampled, where, when and how it defines a mention or citation. Its observations may help with a defined question set, but they should be labelled as estimates or external observations. Do not merge them into Search Console's first-party impression totals[3].

Which Search Console dimensions can I compare?

Google lists pages, countries, dates and devices for the generative AI report. Pages group data by the linked final URL after redirects, with most data assigned to a canonical URL. Countries indicate search origin; dates use Pacific Time; devices distinguish desktop, tablet and mobile. Save the chosen dimension and date range, because a page table and a property-level chart can aggregate differently[1].

AI Overview Tracking in Search Console: What It Shows

Quick Answer: AI overview tracking in Search Console covers impressions for AI Overviews and AI Mode, with page, country, date and device views¹. A US filter measures US-origin exposure rather than worldwide reach, while clicks differ from impressions². Third-party trackers can estimate observed visibility but cannot access Google's internal ranking data³.

Summary in a mind map

AI Overview Tracking in Search Console: What It Shows
│
├─ Define Google's scope
│   ├─ Report combines AI Overviews and AI Mode
│   └─ Impression means a link was shown
│
├─ Choose a dimension
│   ├─ Page, country, date and device are available
│   ├─ Country means search origin
│   └─ Dates use Pacific Time
│
├─ Handle missing data
│   ├─ Small sites may not have enough impressions
│   └─ An absent report is not a measured zero
│
└─ Separate estimates
    ├─ Vendors cannot see internal ranking data
    └─ Keep their sample and method with each estimate

What the Google report includes

Google's current generative AI performance report covers impressions for AI Overviews and AI Mode on Google Search. The default view shows impressions for its supported features. It is therefore a report about a feature group, not a licence to call the whole figure “AI Overview impressions”. Google says it expects the supported list to change over time, so a saved report needs a date and feature description¹.

For a Seattle accounting practice, the useful question might be whether links to its payroll guide appeared in Google's generative results during a defined period. Search Console can answer an exposure question within its scope. It cannot by itself say that a particular customer read the guide, found it helpful or booked a consultation. Those are different observations requiring different evidence¹^,2.

Google excludes Search Labs experiments from the report. This gives another reason to label the coverage precisely. A person who saw an experiment may report seeing the site, yet the business should not expect every such interaction in this report. A discrepancy need not mean that either person fabricated the observation¹.

Impressions and clicks mean different things

An impression in this report is a link to the site shown to a user in a generative AI feature on Google Search. A click is someone clicking a Google link to visit the site. Google's general Search Console definitions distinguish these events. The first is displayed exposure under the report's rules; the second is a visit action from that link¹^,2.

Do not turn one into the other. A payroll guide can receive impressions without clicks. A report of 100 impressions is not a report of 100 visits, nor a count of 100 customers who understood a recommendation. The site owner should place clicks and impressions in separate columns and define the reporting interval in both¹^,2.

Measure	What it records	What it cannot establish alone
Generative AI impression	A site link shown in a supported feature¹	A read, click or business enquiry
Click	A user followed a Google link to the site²	Why that user chose the link
Third-party mention estimate	A vendor's observed or modelled sample	Google's internal ranking data³

The distinction is especially important when a third-party dashboard uses its own “visibility score”. A score may be useful within that dashboard's sampling method, but it is not automatically commensurate with a Search Console impression. Ask for the denominator and collection method before putting the two lines on one chart¹^,3.

Page, country, date and device views

The report offers Pages, Countries, Dates and Devices dimensions. Pages groups data by the final linked URL after redirects, with most performance data assigned to the canonical URL. Countries groups by search origin. Dates can be days, weeks or months and use Pacific Time. Devices separates desktop, tablet and mobile¹.

These details affect a practical export. A redirected service URL may not appear under the address the marketing team remembers. A local team comparing a Monday morning update with a daily figure should account for the report's Pacific Time date boundary. A device breakdown is a view of search-device type, not a profile of the visitor's occupation or intent¹.

Save the selected dimension in the file name or report note. Google says chart totals and table totals can differ because the chart and table use different aggregation. A page-level table should not be presented as if it were the same unit as a property-level total. Retaining the screenshot or export settings makes the comparison auditable¹.

Filter US-origin exposure

The Countries view groups impressions by where the search originated. For a US service business, select the United States and state “US-origin impressions in Google's generative AI report” in the report heading. Do not relabel that filtered number as worldwide generative exposure. Country is a property of the search event, not proof of the searcher's nationality or where the business operates¹.

A firm with clients abroad may also want a worldwide total. It should use a separate view and a separate label. Putting the US and worldwide lines together without the country scope obscures what changed. The question is not merely which number is larger; it is whether the business is comparing the same geography and dates from one report to the next¹.

For related diagnostic work, check whether AI mentions your brand against a fixed question set. That observation answers a different question from the country-filtered Search Console impression count. It should be recorded in a separate evidence column.

When the report is absent

Google says the generative AI report is usually unavailable when a site has not received enough impressions in the supported features. Its help page also points to exclusion from Search generative AI features as a possible reason. An absent tab does not supply a numerical result for the site, and it says nothing by itself about every third-party answer service¹.

A small practice should first confirm it is looking at the correct Search Console property and its inclusion setting. It can note “report unavailable for this property on this date” rather than recording zero. If the report later appears, start a new series with its actual coverage. That preserves the difference between missing data and measured absence¹.

Report state	Defensible record	Avoid saying
Report visible	Export the selected view and period¹	Every impression was an AI Overview
Report absent	Record property, date and setting¹	The site had zero exposure everywhere
External tracker result	Save its sample and method³	It accessed Google's private ranking data

Check third-party estimates

Google says third-party SEO tools lack access to its internal ranking data. It also says their predictions are their own and cannot guarantee performance. This does not make every outside observation worthless. It means the owner should ask what was actually observed and which steps were estimated³.

A useful vendor record names the questions tested, search interface, geography, dates, devices, repeat schedule and definition of a citation or mention. If the tool claims to report “internal Google metrics”, request its source and a precise definition. A visible result sampled by a vendor is not an internal ranking signal. Keep that distinction in the purchase decision and in reports to clients³.

Whether ChatGPT recommends your business and why an AI service may recommend a competitor are useful follow-up questions. They call for observed answer evidence rather than renaming Google's combined report.

Report the source, dimension and geography before comparing numbers.

In words, the tree branches like this: if the Google report is available, choose its feature scope and dimensions; filter US-origin impressions for the US question; if the report is absent, record missing data; keep vendor estimates in a separate column with their method.

What the numbers cannot decide

The cited report does not give a guarantee that a particular prompt will show the site tomorrow. An impression says a link was shown; it does not certify a stable answer position or future selection. A vendor's forecast adds another layer of uncertainty. Both kinds of evidence belong at their own level. The wider question of generative engine optimisation needs a separate assessment¹^,2^,3.

What to do next

Export one defined period for the practice's priority guide, preserve the chosen page and country settings, and compare the next period on the same basis. Keep clicks separate from impressions and document the sample behind any outside estimate. Use AI Answer Audit for a one-time check of how AI answers describe and recommend the business.