Google Search Console Generative AI Report: Measure AI Overview Visibility

What the Search Console generative AI report measures

Google introduced a dedicated generative AI performance report for Google Search. It is currently rolling out to a subset of website owners, and a property may also need enough impressions before the report appears. The report covers organic impressions from AI Overviews and AI Mode; Search Labs experiments are excluded.

An impression means a link to the site was shown in a supported generative AI feature. The report can group that visibility by canonical page, country, date, or device. It is useful first-party Google data, but it is not a query-level citation tracker and it does not describe visibility in non-Google AI products.

Use the generative AI and Performance reports together

The new report answers one narrow, valuable question: which pages are receiving impressions in Google's supported generative AI features? The standard Search results Performance report remains the place to analyze queries, clicks, impressions, CTR, and average position for ordinary Web search data.

Use the page dimension as the bridge. A page with generative AI impressions can be reviewed against its standard search queries and performance. A page with proven query demand but few or no generative AI impressions becomes a research candidate—not proof that an AI feature caused lost clicks.

• Generative AI report: AI Overview and AI Mode impressions by page and context.
• Performance report: query, click, impression, CTR, and average-position evidence.
• Page review: relevance, originality, factual support, crawlability, and user experience.
• Citation checks: separate snapshots for supported non-Google grounded answer providers.
• Do not infer a direct AI Overview loss from low CTR alone.

The five Search Console signals CiteWins uses

Search Console exports can be noisy. A site can have thousands of query rows, many of which are too small to act on. The key is to turn the export into a shortlist of page-level decisions.

CiteWins groups raw query and page data into five opportunity types. Each type has a different content response, so the action plan does not treat every issue as another blog rewrite.

A step-by-step GSC to GEO process

The process starts with evidence and ends with a published improvement. The important part is that every content recommendation should tie back to a query, a page, and a measurable outcome.

That is where many manual audits break down. They produce advice like 'improve structure' without saying which page, which query, which section, or what to measure after publishing.

1. 1

   Export or sync the last 90 days of query, page, click, impression, CTR, and position data.

2. 2

   Group data by page and isolate the queries with meaningful impressions.

3. 3

   Compare pages in the generative AI report with the queries and performance visible in the standard report.

4. 4

   Inspect the page for technical eligibility, intent match, original value, evidence, usability, and relevant internal links.

5. 5

   Publish a targeted fix instead of rewriting the whole page.

6. 6

   Measure a before-and-after window so the team can see whether the fix moved clicks, CTR, or citation status.

How to prioritize when everything looks important

The highest-traffic page is not always the best first fix. A page may have high impressions but be too broad, too competitive, or already near its realistic ceiling. GEO prioritization needs both upside and feasibility.

A practical scoring model looks at search demand, current position, CTR gap, page quality, and how much work the fix requires. That turns a long report into a backlog.

• Start with pages already earning impressions for buyer-intent queries.
• Favor pages where the answer can be improved without a full rewrite.
• Prioritize pages that support revenue, signups, demos, or lead capture.
• Avoid spending the first sprint on low-impression informational scraps.
• Keep a measurement window so wins can be proven.

What Google's current generative AI guidance changes

Google's July 10, 2026 guidance says its generative AI features are rooted in core Search ranking and quality systems. From Google's perspective, AEO and GEO work for Google Search is still SEO. The durable priorities are crawlability, index eligibility, a good page experience, and unique, expert, non-commodity content that satisfies visitors.

Google also says there is no special AI schema, no need to rewrite content into a particular AI style, and no ranking benefit from llms.txt in Google Search. Structured data remains useful when it matches visible content and supports an existing Search feature; it should not be added as an invented GEO shortcut.

• Create original evidence, first-hand detail, and a useful point of view.
• Keep important content crawlable, indexable, and available as text.
• Use headings and sections for readers, not a rigid extraction formula.
• Treat llms.txt as optional for other systems; Google Search ignores it.
• Measure Google's generative AI visibility in its dedicated Search Console report.

What if you do not have Search Console yet?

You can still start. A public website scan can inspect crawlability, indexability hints, titles, meta descriptions, headings, schema, sitemap discovery, and AI-readiness issues. It will not know your private query data, but it can find public problems that should be fixed before GSC data becomes useful.

Once GSC is verified, collect data for at least a few days. New and low-traffic sites may need several weeks before the query set is useful enough for prioritization.

Primary sources

Official documentation used to review and update this guide.