Search Console AI Mode data is available in Google Search Console’s Generative AI performance report, which shows impressions for links in AI Overviews, AI Mode and generative AI cards in Discover. It does not show the prompts that triggered those links, clicks, click-through rate or position. The AI report is a filtered view of data already included in the regular Web performance report, so adding the two impression totals together will double-count.
This guide explains how to find the report, what each metric means, how AI impressions blend into ordinary search reporting, and how to use the numbers without claiming more than they can prove. If you need to compare this data with a wider AI visibility programme, SearchFIT’s resources and pricing explain how its tracking and reporting fit into a broader workflow.
What Search Console AI Mode data actually shows
Google’s dedicated report is called the Generative AI performance report. You can find it in Search Console under Performance → Generative AI (Search), if the report is available for your property. Its central metric is impressions: instances where a link to a page on your site appeared in a supported generative AI search experience.
The report combines several surfaces. It includes links shown in AI Overviews, Google’s AI-generated summaries within Search, and AI Mode, Google’s conversational search experience. The reporting also covers generative AI cards in Google Discover. Search Labs experiments are excluded from production reporting, according to Google’s report documentation.
An impression is not the same thing as a click, a visit or a person reading your content. It means Google counted a link as having been shown under its impression rules. A page can earn impressions without receiving any traffic from that appearance.
The report groups results by four dimensions:
| Dimension | What it tells you | A useful question |
|---|---|---|
| Pages | Which landing page URLs received AI impressions | Which of our pages are appearing as sources? |
| Countries | Where those impressions were recorded geographically | Does visibility differ across markets? |
| Devices | Whether impressions came from desktop, mobile or tablet | Is our visibility concentrated on a device type? |
| Dates | How impressions changed over time | Did a page or market change after an update? |
A dimension is a way to break a metric into groups. For example, the impression total is the metric; Pages is the dimension that lets you see which URLs contributed to it. The report does not provide a separate filter for AI Overviews versus AI Mode, so you cannot use it to compare those two surfaces directly.
The report also does not expose the prompt or search query behind each impression. You cannot look at a row and see the exact question a person asked before Google cited your page. Google’s report documentation describes the available data and dimensions, while its Search generative AI control documentation explains the separate setting that controls eligibility.
What the report does not show
It is just as important to understand the missing data as the visible data. The dedicated Generative AI report does not include:
- Queries or prompts that caused a page to appear.
- Clicks from links shown in AI experiences.
- Click-through rate, usually shortened to CTR.
- Average position.
- A breakdown that separates AI Overviews from AI Mode.
CTR means the share of impressions that resulted in clicks. Since the dedicated report has no click count, it cannot calculate a direct AI CTR. Do not divide clicks from the standard Web report by AI impressions and label the result AI CTR. The clicks in the Web report are blended with other search activity and are not identifiable as AI clicks there.
Google announced the dedicated reporting in June 2026 and completed the worldwide rollout to qualifying sites on August 31, 2026, according to its Search Central announcement and Search Console Help. The report may not appear for every property. Google says a property needs a minimum level of impressions in generative features before the report becomes available. Search Labs activity is also excluded, and a property set to exclude generative AI features will not earn those impressions.
Search Console AI Mode data and regular Web data
The main reporting trap is treating the Generative AI report and the Web report as separate piles of impressions. They are not. The Generative AI report is a filtered sub-view of eligible data that is already included in the standard Performance → Search results report when its search type is Web.
Think of the Web report as the wider set. It includes the ordinary organic results and also includes eligible impressions and clicks associated with AI search features. The Generative AI report extracts a subset of that search activity to show AI impressions. It is useful for identifying URLs with generative AI visibility, but it is not an extra source of impressions to add on top.
For example, suppose a property has 10,000 Web impressions in a selected period, and its Generative AI report shows 1,200 impressions for the same period and scope. Do not report 11,200 combined impressions. The 1,200 are already within the wider Web total. The example numbers are illustrative, not a Google benchmark.
Google’s documentation on impressions, position and clicks describes how Search Console metrics work. The special challenge here is that the Web report retains its usual metrics while AI activity is folded into those totals. This means Web clicks and positions can be affected by AI appearances, but the report does not label each row as an AI interaction.
Why impressions are not additive
Use the Generative AI report to answer a question such as, “Which of our pages are appearing in generative search results?” Use the Web report to understand overall search performance, including clicks and impressions across the broader set of Web results.
Do not add the two impression totals. That would count the generative AI subset once in the Web total and again in the dedicated report. Instead, report the numbers side by side and label their scopes clearly:
- Web impressions: the broader Search Console Web performance total.
- Generative AI impressions: a filtered view of eligible appearances in supported generative AI surfaces.
For dashboards, use separate tiles or columns rather than a combined total. Add a short note that the AI figure is included in the Web figure. This simple annotation prevents an apparent increase in total visibility from being mistaken for growth when it is only two views of overlapping data.
Why one page can contribute just one impression
Search Console applies de-duplication rules. If the same URL appears as an AI Overview citation and also as a standard organic listing for the same query, it does not necessarily receive two impressions in the aggregate Web report. The reporting logic counts an impression for that URL in the context of the result, rather than simply adding one for every visual location where the link appears.
This matters when explaining differences between what you see on a search results page and what the aggregate report says. A page that appears in more than one part of the experience may not produce a matching increase in aggregate impressions. Google’s Search Console metric guidance and reporting explanations are the best reference for interpreting the official totals; avoid reverse-engineering exact exposure counts from a screenshot.
What counts as visible
A link can be visible in an AI experience without a person scrolling to it. Under the stated impression rules, links rendered in a visible part of an AI Overview can count even if the user does not physically scroll down to that part of the page. A citation concealed behind a control such as “Show more”, or behind a carousel item that has not been revealed, does not count until it becomes visible through interaction.
That distinction means impressions are not a measure of attention. A link might technically be on screen but receive little notice. Conversely, a source hidden behind an expansion could be relevant to the generated answer but not count until the user opens it. Search Console impressions should be interpreted as a visibility signal under Google’s rules, not proof that a user read or remembered a citation.
How position works in the Web report
The dedicated Generative AI report has no position metric. However, AI appearances can affect average position in the standard Web report. In an AI Overview, links in the same block can inherit the position assigned to that block. If Google records an AI Overview at position 1, multiple cited URLs can be reported at that same position, even though one citation appears more prominent than another.
Position in Search Console is not a simple measure of where a link sits on a user’s screen. A position of 1 does not mean a page was the most prominent citation, and the same position for several URLs does not mean they were shown in identical ways. In September 2026, Google’s John Mueller clarified that links in an AI Overview block share a position slot, while also noting the complexity of reporting position in conversational AI Mode. Search Engine Watch has reported on how AI Overview impressions are counted.
This makes average position harder to interpret when AI features appear. It is an average across recorded appearances and result types, not a rank tracker’s exact view of each URL in a stable list. Avoid saying that a page “ranked first in AI” solely because its average Web position is 1.0. The AI report itself cannot tell you that.
How to read the report without overclaiming
Start with a clear question. The data supports some questions well, some only indirectly, and some not at all.
| Question | Can the report answer it? | Best way to use the data |
|---|---|---|
| Which URLs receive AI impressions? | Yes | Review the Pages dimension and compare periods. |
| Which country or device has more recorded impressions? | Yes | Use the Countries and Devices dimensions. |
| Did AI visibility rise or fall? | Yes, at the impression level | Compare dates, while checking for data changes and seasonality. |
| Which exact prompts triggered a citation? | No | Use page topics and other query research to form hypotheses, not claims. |
| How many AI citations produced clicks? | No | Use broader Web clicks as context, not as AI-attributed clicks. |
| Was an impression from AI Mode or an AI Overview? | No | The report combines those surfaces. |
Step 1: Confirm the report and property scope
Open Search Console and select the correct property. A property is the website or website section whose data Search Console is collecting. Check the report’s date range, country and device context before comparing totals. If the Generative AI report is missing, first check whether the property has enough qualifying activity and whether its generative AI setting is set to include.
Google’s Search generative AI control is a setting for whether a site’s content and links can be used for grounding and citations in covered AI experiences. “Grounding” means using sources to inform an AI-generated response. Google says the default is Include. If a property is set to Exclude, generative AI impressions can drop to zero because its content is no longer eligible for those appearances.
The control is not a way to separate AI traffic in analytics. It changes eligibility, not the report’s available metrics. Before interpreting a sudden change, record whether the setting changed and when. Also check for platform or reporting notices. For example, Google documented a logging anomaly affecting Generative AI impression counts from August 13 to August 21, 2026, and said it had been resolved. A reporting anomaly can make a trend look like a change in user behaviour when it is actually a data issue.
Step 2: Find the URLs doing the work
Open the Pages dimension and export the URLs with impressions. Start by grouping pages into meaningful types: educational guides, category pages, product pages, product detail pages, comparison pages and support content. This segmentation helps you identify what kinds of pages appear, rather than treating every URL as unrelated.
Look at both the absolute number of impressions and the change over time. A page with a small total may have grown sharply from a very low baseline. Another page may lead the site in impressions but have little recent movement. Neither observation alone proves that its content is better or worse. Check the page’s topic, search demand, seasonality, indexability and the date of any content change before deciding what caused the pattern.
For an e-commerce brand, a useful review might compare a category guide about choosing a product with product pages containing specifications, availability and pricing. If the guide receives impressions and the product pages do not, that may suggest the guide answers common informational questions more clearly. It does not prove that Google’s AI system lacks product information, or that a technical issue is the cause. Inspect the pages and their search performance before making that diagnosis.
For a B2B SaaS company, compare solution pages, integration pages, pricing explanations and educational articles. A strong impression trend on an explainer can show that the page is appearing in generative search contexts. It cannot tell you whether those impressions came from prospects evaluating vendors, students doing research or another audience. Use the finding as a prompt to review the content and validate it through other evidence.
Step 3: Compare AI impressions with Web impressions carefully
A useful internal comparison is the AI citation ratio: Generative AI impressions for a URL divided by Web impressions for that URL in the same time period and property scope. It estimates the share of the URL’s reported Web impressions represented by the AI subset. It is a diagnostic ratio, not an official Google metric and not a measure of clicks or audience share.
For example, if a page has 400 Generative AI impressions and 2,000 Web impressions in the same period, the ratio is 20%. Those figures are illustrative. Make sure both totals use comparable dates and filters, and remember that Google’s data can be affected by aggregation and privacy thresholds. A high ratio may indicate that a page’s visibility is relatively concentrated in generative features. It does not mean the page receives 20% of its visitors from AI.
Use the ratio to find pages worth reviewing, not to set a universal pass mark. A glossary page may reasonably have a different pattern from a product page. A ratio above a chosen threshold, such as 40%, can be used as an internal flag if your team defines it that way, but it is not a Google standard or proof of content quality. SearchFIT’s AI search visibility resources can support a wider measurement approach, but keep Search Console’s own metrics clearly labelled in client reporting.
Step 4: Use Web queries as clues, not prompt attribution
Because the Generative AI report omits queries, marketers often look at the Web report’s query dimension for possible clues. You can export pages from the AI report, then inspect the queries associated with those same URLs in the standard Search results report. Long, specific or conversational queries may help you understand the page’s broader search context.
This is a way to form hypotheses, not to recover the prompt behind an AI citation. A query in the Web report may have led to a standard result, an AI Overview, AI Mode activity or another eligible search interaction. There is no reliable row-level connection that proves a particular query generated a particular AI impression. Do not tell a client that a prompt triggered a citation unless you have a separate, valid source for that claim.
Multi-turn AI Mode sessions add another complication. Follow-up phrases, including conversational responses, can appear in ordinary query reporting, but the Generative AI report does not connect those terms to a specific URL impression. Google had not announced query-level or direct click attribution data for these AI features as of November 2, 2026, according to Search Engine Watch’s coverage. Keep the distinction clear: query data can offer context, but not attribution.
Step 5: Read changes alongside clicks and conversions
If AI impressions rise while Web clicks fall, the reports do not establish that AI caused the decline. Other factors may include changes in ranking, demand, seasonality, competition, site changes or the mix of queries. Compare equivalent periods, review the standard Web report and check your analytics and conversion data. Treat the relationship as a signal to investigate, not a causal result.
Some research has found that AI Overviews can change click behaviour, especially for informational searches where a summary may satisfy the searcher’s need on the results page. However, do not apply a general click-through decline to your site without evidence. A study reported by Search Engine Journal described Google’s AI search reporting as inadequate for direct attribution, and a separate opt-out study discussed by Search Engine Roundtable derived an estimated AI Overview CTR of about 0.9% in its specific case study. That is a case-specific estimate, not a reliable benchmark for your pages or industry.
The same opt-out study described a possible “SERP halo effect”: being cited in an AI Overview may support clicks to a brand’s ordinary organic listing. That finding does not mean every AI citation increases clicks. It is a reason to evaluate organic performance and AI visibility together, rather than assuming the dedicated AI report alone can measure the business effect.
A practical reporting workflow for teams
A repeatable workflow helps agencies and in-house marketers avoid inconsistent interpretations. Keep the reporting light enough to run regularly, but specific enough to connect the observations to decisions.
- Choose the period and scope. Record the property, dates, country and device filters. Compare like with like.
- Export the Generative AI report. Save the page, country, device and date breakdowns, including the export date.
- Export the Web report separately. Include impressions, clicks, CTR, position and queries where useful. Label these as broad Web metrics, not AI-attributed outcomes.
- Join by URL for analysis. Compare AI impressions with Web impressions for the same page and period. Do not sum them.
- Group URLs by page type. Separate informational content from commercial and transactional pages.
- Annotate changes. Record major content updates, technical releases, setting changes and known data anomalies.
- Make one testable decision. For example, improve a product page’s accessible specifications, or add original evidence to a guide that attracts impressions but offers little reason to visit.
- Review again after a suitable interval. Compare the same measures and acknowledge that other changes may have influenced the outcome.
When sharing results with a client or executive, use phrasing such as: “This URL received more recorded impressions in Google’s combined generative AI report this month.” Avoid statements such as: “This page earned 100 AI clicks,” “These were the prompts,” or “We ranked first in AI,” unless another measurement source actually supports them.
For agency reporting, separate observed data from interpretation. A clear report might show the AI impression trend, a list of URLs with the largest changes, broader Web clicks and CTR in a separate panel, and a short note on limitations. SearchFIT’s comparison page and pricing information can help teams decide how to complement Search Console with ongoing AI visibility tracking. No third-party visibility measure should be presented as a replacement for Google’s own Search Console data; each tool measures a different view of search presence.
What to do with the findings
Use AI impressions to prioritise content review. If an informational page repeatedly appears, check that it answers its subject clearly, is accurate and gives readers a useful next step beyond the basic answer. If its content is easily summarised and offers no distinct value, consider adding original research, first-hand expertise, a practical template or a tool. These additions can improve the page for people, but they are not a guarantee of AI citations or clicks.
If a commercial page receives few or no AI impressions, inspect it before assuming the cause. Confirm that the page is indexable, important details are available in the rendered page, product or service information is clear, and the page explains what the offer does and who it is for. A page using JavaScript is not automatically excluded, and an absence of impressions alone does not prove a rendering problem. Check technical evidence, the page’s ordinary search performance and the content itself.
For e-commerce, product facts such as variants, specifications, stock information and delivery terms should be clear to customers and consistently represented. For SaaS, clarify the product category, use cases, integrations, limitations and pricing context. These are sound content and usability practices. Search Console’s AI report can tell you whether some pages received impressions, but it cannot diagnose the exact reason a model cited one source and not another.
Do not opt out simply because the report cannot show direct clicks. Google’s control affects whether the site is eligible for covered AI features; exclusion is not a way to make competitors disappear from answers. The cited opt-out case study reported a drop in ordinary organic clicks after exclusion, but that result should not be treated as universal. Assess your own business goals and use the setting only after understanding its implications. Google’s control documentation is the authoritative place to check the current setting and its behaviour.
The simplest way to explain the data
If you need to explain Search Console AI Mode data in one minute, say this: Google now reports impressions for links appearing in a combined set of generative AI experiences. The report helps identify pages, countries, devices and trends, but it does not show prompts, clicks, CTR, position or a split between AI Mode and AI Overviews. Those impressions are already included in the broader Web report, so show the two figures side by side and never add them together.
That framing is both useful and honest. It lets you act on a new visibility signal while respecting what the data cannot prove. Use the AI report to discover where your content appears, the Web report to understand broader organic performance, and your analytics and business data to evaluate outcomes. For a next step, explore SearchFIT’s resources and pricing to see how AI visibility tracking can sit alongside your Search Console workflow.