Google Search Console now has a dedicated generative AI performance report for Search. Google says the report includes impressions from AI Overviews and AI Mode, with pages, countries, devices, and dates as dimensions; it does not provide a separate click metric in this report. The report is not a dashboard for ChatGPT, Perplexity, or other non-Google assistants. Its insights rolled out worldwide on August 31, 2026. Google Search Console Help: Generative AI performance report · Google Search Central announcement
Use it to see which pages receive Google generative-search impressions and how those impressions vary by country, device, or date. Do not treat impressions as visits, citations across all AI platforms, or evidence of a business outcome. The report is an additional view within Search Console’s performance reporting, and its aggregation and data limits matter when you compare totals.
Confirm the report and its scope
Open the Search Console property you actually own and select the Generative AI performance report for Search. The report covers AI Overviews and AI Mode on Google Search; Google Discover has a separate report. Search Labs experiments are excluded. If the report is absent, Google says the site may not yet have enough impressions in these features, or it may have been excluded. Check property access and eligibility before inferring that the site has no AI visibility.
Read the available metrics without inventing clicks
The Search report shows impressions: how often links to the site were shown in a generative AI feature. It supports grouping by final page URL, country, date, and device, and filtering by search type such as text or multimodal. Dates in this report use Pacific Time. Some newest data may be preliminary and can change. Google describes chart and table aggregation differently; chart totals can differ from table totals, especially when one groups by property and the other by page. In exported data, values displayed as ~ or - may export as zero. Preserve those reporting details in any analysis.
Do not divide impressions by clicks from another report and label the result a report-native CTR unless the metric definitions and populations actually align. The report helps answer “which pages appeared and where?” It does not independently tell you whether a person visited, cited the page in another assistant, or converted.
Build a repeatable review around a decision
Start with a business question: Which Arabic or English pages are appearing for audiences in a target country? Did a known content update coincide with a change in impression distribution? Is one high-priority page absent from the report’s page table? Record the reporting period, property, filters, selected dimensions, date time zone, and export date. Compare like with like and annotate page releases, seasonality, or major site changes.
Use the data to choose a page for a human quality review: does it answer the audience task clearly, contain accurate and distinctive evidence, load and render for Search, and link to a relevant next step? An impression change alone does not show why a system selected a page. Avoid creating pages solely to multiply queries or impressions; Google’s current guidance continues to emphasize useful, original content and foundational SEO.
Short-video script: What the new Search Console report can tell you
Hook: “Seeing your site in Google’s AI Search report? That number is impressions, not visits.”
Beat 1: “The report covers AI Overviews and AI Mode in Google Search—not every AI assistant.”
Beat 2: “Review pages, countries, devices, and dates; check time zone and preliminary data.”
Beat 3: “Chart and table totals can differ because aggregation differs.”
Close: “Use it to choose a page to review; verify clicks, visits, and outcomes in their own reports.”
Educational carousel: Read the Search Console generative AI report
- Cover: “New report, new visibility signal. What does it measure?”
- Scope: AI Overviews and AI Mode in Google Search; Discover is separate.
- Core measure: impressions, not a report-specific click count.
- Dimensions: pages, country, device, and date; search type can be filtered.
- Date caveat: report dates use Pacific Time and new data may be preliminary.
- Aggregation caveat: chart and table totals may differ.
- Absence caveat: low impression volume or exclusion can explain a missing report.
- Next step: review the page and validate business outcomes elsewhere.
Field experiment thread: Test one page improvement against an observation window
Proposed monitoring plan, not proof of ranking or causal impact.
1/ Select one important page and define the audience, language, country, and task it serves.
2/ Export a baseline from the Search Console generative AI Search report. Save property, filters, dimensions, period, and export date.
3/ Record impressions by page, country, and device where available. Note Pacific Time and any preliminary dates.
4/ Review the page for accuracy, original evidence, useful structure, and technical accessibility. Make only a documented, user-serving change.
5/ Keep a log of the change date and other site releases, seasonality, or campaigns that may affect the period.
6/ After a comparable observation window, export the same dimensions and compare impressions; do not imply visits or conversions from impressions.
7/ Mark the result as observed, mixed, or inconclusive. Validate visits and business outcomes in their own systems and avoid claiming the edit caused a change without a suitable design.
Tool-neutral prompt pack: Analyze an exported Search Console AI report
Report map: “Using the attached export and report settings [file], summarize impressions by page, country, device, and period. Preserve the timezone, filter, property, and aggregation notes. Flag preliminary dates and ~ or - values converted to zero.”
Change review: “Compare these two periods for the same property and dimensions [exports]. Identify changes in page/country/device impressions, list known site changes, and separate observation from possible explanation. Do not infer clicks, citations outside Google, or conversions.”
Page review: “Given the page [URL/text] and its Search Console impression pattern [data], assess whether it clearly satisfies the stated user task, provides accurate distinctive evidence, and is technically accessible. Do not recommend query-specific pages merely to inflate impressions.”
Stop-doing swipe file: Replace Search Console AI report myths
Stop: “AI impressions mean visits.” Use: “The report counts appearances; check visits in analytics.”
Stop: “This report shows ChatGPT and Perplexity citations.” Use: “It reports Google Search generative AI features; measure other platforms separately.”
Stop: “No report means Google excluded our whole site.” Use: “Check eligibility, impression volume, and whether the property was excluded.”
Stop: “The chart and table disagree, so the export is broken.” Use: “Compare property-level and page-level aggregation, filters, and time periods.”
Stop: “An increase proves our page change caused it.” Use: “Record the change and confounders; treat the movement as an observation until causality is tested.”
Hook–Value–Template: Turn the report into a reviewable decision
Hook: “Google’s new AI visibility report measures appearances—not whether people visited or converted.”
Value: A repeatable export by page, country, device, and period helps choose what to review while preserving the reporting caveats.
Copy-ready review: “Property: [property]. Report: [Search generative AI]. Period and PT dates: [range]. Filters/type: [details]. Dimensions: [pages/countries/devices]. Impressions: [values]. Preliminary or missing values: [notes]. Chart/table aggregation: [scope]. Page change: [date and user-serving reason]. Other changes: [list]. Observation: [what moved]. Unknown: [clicks, causal reason, other-AI citations, conversions]. Next check: [page review or separate measurement].”