Validating Search Console Bulk Export Partitions and Retries
When Search Console sends performance data to BigQuery, it writes each day's data into date-based partitions, which keep each date's rows together.…
Website owners need to know whether people see their pages in Google AI Mode or AI Overviews, two Google Search features that show AI-generated answers. That distinction matters when you measure how often people see or visit your site. Google’s published Search Console reports do not reliably separate the two features.
Search Console’s Generative AI performance report counts combined impressions, or times pages appear, for AI Mode and AI Overviews. The standard Performance report includes activity from both under the Web search type. Website analytics and server logs start recording activity after a visitor reaches your site. Keep these measures separate. Label findings about a specific feature as manual observations unless Google provides a supported way to identify it.
AI Overviews are generated summaries within Google Search results, and link to web pages that support the answer. Google shows an overview when its systems determine that it would add value to the search results.
AI Mode is an interactive Search experience for queries that call for more exploration, reasoning, or complex comparisons. Both features can use “query fan-out,” which means Google searches related subtopics or sources to build an answer. The answers and links can differ. Google describes these features and their inclusion in Search in its documentation on AI features and your website.
Google’s documented reports for site owners do not separate the two features. Describe them separately when reporting what you observed, but do not assign combined Search Console data to either feature.
The Generative AI performance report is Google’s closest built-in measure of visibility in these features. It counts impressions, meaning times pages appear in Search, and lets you group results by page, country, date, or device and filter by search type. Its documented fields do not let you choose AI Mode or AI Overviews separately. The report also gives no separate click count, click-through rate (the share of impressions that result in clicks), or position for either feature. Google says the report can be absent when a property has not received enough relevant impressions. A missing report does not mean the site had no visibility. Check the report’s metrics and filters before drawing conclusions from an empty or missing view. The Search Analytics API lets you request Search Console search-performance data; the table below summarizes what it and other sources can show.
| Evidence source | What it can show | What it cannot establish by itself |
|---|---|---|
| Search Console Generative AI report | Combined impressions, grouped by the report’s available dimensions and filters | Separate AI Mode and AI Overview totals, surface-specific clicks, or exact prompts |
| Search Console Performance report | Web clicks, impressions, click-through rate, and average position, including AI-feature activity | Which activity came from AI Mode, AI Overviews, or other Web results |
| Search Analytics API | Search-performance data using available dimensions and filters, including searchAppearance |
A guaranteed, documented classification that separates the two AI surfaces |
| Google Analytics and server logs | Visits or requests that reached the site, landing pages, and on-site activity, subject to your setup | Search impressions or a guaranteed source value that distinguishes the two surfaces |
Google’s Performance report methodology says clicks on external links in either feature count as clicks, and AI-feature activity appears under Web. It also explains how Google counts position, or where a result appears: links in an AI Overview share the overview’s position. Average position measures Web results and does not rank AI Mode or AI Overviews.
The Search Analytics API supports searchAppearance to filter or group results by how a result appeared. Google advises querying the API to see which values are available for a property. This feature does not guarantee separate values for AI Mode and AI Overviews. Do not assign a feature label to an API value unless Google documents what it means and your data returns it. See the Search Analytics query reference.
Analytics and server logs begin measuring after someone reaches your site. They cannot count people who saw an answer but did not click through. Google does not guarantee that visits from either feature carry a consistent referrer (information about where a visit came from) or traffic-source value that identifies the feature. Use these tools to report visits according to the source definitions they record, not to guess which feature sent the visitor. Google explains why Search Console clicks and Analytics sessions differ.
Use separate labels for separate measures. Call the Generative AI report metric combined AI-feature impressions. Call clicks from the ordinary Performance report Web clicks unless a supported dimension identifies a narrower source. Call Analytics activity organic-search sessions only when that label matches the property’s channel and source definitions.
Keep these measures separate because they count different events. An impression records that a page appeared in Search under Google’s reporting rules. A click records someone selecting a result link, and a session records activity on your site. If combined AI-feature impressions and sessions both rise, that does not show that AI Mode or AI Overviews caused the increase.
Exporting data does not add missing categories. Google’s report export contains the chart and table data available in the report, but it does not separate AI Mode from AI Overviews. The same applies to API queries: they return available fields, but cannot provide a distinction that Google does not document.
To track a specific feature, record what you see in individual searches. Record the query and date for each search, then note the device and market. Label each result as an appearance you observed under those conditions. These samples document your observations; they do not give a complete impression count or, by default, a representative estimate. The Generative AI report does not provide prompt-level data. For repeatable observation records, see how to check a site in AI Overviews and AI Mode.
When you compare periods, use the same property, date range, device, country, and search-type filters. Search Console uses Pacific Time, and its data can be delayed or preliminary. Search Console and Analytics may use different time zones and measurement rules, so their totals may not match.
Example data (article setup: technical explainer: direct mechanism; private trial 2026-10-04). Hypothetical illustration only, not Google product data or a claim about an actual test.
| Metric | Example value | Safe label |
|---|---|---|
| Generative AI report impressions | 1,000 | Combined AI-feature impressions: AI Overviews and AI Mode |
| AI Overview impressions | Not separately reported in this view | No separate native figure |
| AI Mode impressions | Not separately reported in this view | No separate native figure |
| Search Console Performance report clicks | 42 | Web clicks; surface not identified |
| Analytics sessions from Google organic | 38 | Sessions under the property’s stated Analytics definition |
The example shows how to label the figures that are available. It does not show how to divide 1,000 impressions between AI Mode and AI Overviews. Do not assign part of the combined total to either feature unless you use a separate, transparent estimation method and explain its limits.
Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.
Take a look at vroni.com