HOW TO MEASURE AI VİSİBİLİTY: THE SEARCH CONSOLE AI REPORT AND GA4 SETUP

You cannot measure visibility in AI search from a single screen, because the data lands in two separate places. Google Search Console's generative AI report gives you impressions only, while the click side becomes visible in GA4 once you isolate the chatgpt.com and perplexity.ai referrers. Accurate measurement starts by joining these two half data sets in one dashboard. Below you will see exactly which metric lives where, who gets access to the report, and how to recover the click side that the report leaves out.

Why measuring AI visibility differs from classic SEO

The whole difference fits in one sentence: AI surfaces leave the impression in one place and the click in a completely different one. In classic SEO, impressions, clicks, click-through rate and position all sit side by side in a single report. With generative AI features, Google tells you only how many times you were shown; whether anyone actually reached your site has to be read from analytics.

That split changes the logic of measurement. Instead of hunting for a single "visibility score," you answer two questions separately. Am I being shown? belongs to Search Console. Is that visibility turning into traffic? belongs to GA4. Dashboards that mix the two produce wrong conclusions, because they count two different halves of the same event.

If you are after the measurement of visibility rather than the tactics of earning it, you are in the right place. We covered the methods for getting your brand into answers separately; for those, see our guide on how to become visible in AI search. Here the single subject is turning that effort into a number.

What does the Search Console generative AI report show?

The report gives you one metric: impressions. In Google's own words, an impression is "how many times links to your site were shown to a user in a generative AI feature on Google Search." Clicks, click-through rate and average position do not appear in this report. The signal in your hands answers only the question of how many times you were listed.

You can group the data by four dimensions:

• Pages: Groups data by the final URL linked by a generative AI feature after any redirects.
• Countries: Groups data by the country where the search originated.
• Dates: Groups data by days, weeks or months based on the selected time granularity.
• Devices: Splits data by desktop, tablet or mobile.

The point to notice is that there is no Query dimension in this list. You cannot learn from this screen which questions brought you in. In the same way, the "Search appearance" dimension from the classic Performance report is absent here. You are left with page, country, date and device; you read your content strategy across those four axes.

What is missing: clicks, CTR, position and queries

There are four areas where the report stays silent, and each opens a gap in measurement. No clicks, no click-through rate, no average position, no queries. Any reading of visibility that ignores them is only half finished.

Here is what those gaps mean in practice:

• Without clicks you cannot tell whether impressions turned into traffic; a high impression count on its own is not success.
• Without position you cannot measure where you ranked inside the answer; on AI surfaces the classic notion of "rank" blurs anyway.
• Without queries you cannot know which intent brought you in, so you have to infer content priority at the page level.

For contrast: the click and conversion metrics used in standard success measurement already exist in the classic Search performance report as clicks, impressions, CTR and average position. The generative AI report deliberately carries only the impression leg of that set. You rebuild the missing three legs through GA4.

Who can access the report?

The report is not open to everyone; Google is serving it to "a subset of website owners." In the document's wording, the feature is being rolled out to a portion of site owners first, to allow thorough testing before wider release. If you do not see the report in your Search Console at all, the fault may not be yours.

Google lists two typical reasons. If your site has not received enough impressions in generative AI features, the report will not show. If your site was excluded from those features, no data lands either. So the absence of the report sometimes means "not your turn yet" and sometimes "not enough volume." If you have no access, you have to base your measurement plan temporarily on GA4 referral data alone.

Seeing the click side in GA4: chatgpt.com and perplexity.ai referrers

The click side appears in GA4's Traffic acquisition report as a referral source. When a user clicks a link inside an AI answer and reaches your site, the session source becomes that tool's domain and the medium becomes referral. By GA4's definition, referral is the medium assigned to traffic arriving through a non-ad link, and the source is the sending domain.

In practice the two referrers you look for are:

• chatgpt.com and its variants (in some setups chat.openai.com): clicks coming from inside ChatGPT.
• perplexity.ai: clicks from the citation links in Perplexity answers.

You isolate these by filtering on the session source/medium dimension in GA4. Knowing how each source behaves is easier once you understand the logic of showing up in ChatGPT and appearing in Perplexity, because the two engines generate citations differently, so their click behavior differs too. On the Google AI Overviews side there is no separate referral domain; that traffic comes from within google.com, and separating it stays tied to the Search Console generative AI report. We explained what AI Overviews are in our AI Overviews guide.

Joining two half data sets in one dashboard

Accurate measurement means taking the impression half from Search Console and the click half from GA4 and placing them side by side on the same time axis. That way you read "how much am I shown on Google's AI surface" and "how many clicks am I getting from AI tools" in one glance. Since the two sources come from different systems, they cannot be added one to one; the goal is not to sum them but to align their trends.

When building the dashboard, keep this distinction:

• Search Console impressions = how often you are listed on generative surfaces inside Google Search. Google only.
• GA4 referral clicks = actual visits reaching your site from the answers of tools like ChatGPT and Perplexity. Also captures non-Google engines.

Because the two metrics cover different engine universes, they complement each other rather than confirm each other. Impressions on Google can fall while Perplexity clicks rise; that is not a contradiction but news from two separate channels. The dashboard's value is that it makes both channels readable on the same calendar.

Impressions or clicks? What each metric tells you

Impressions measure reach and clicks measure interest, so treating one as a stand-in for the other is a mistake, because they answer different questions. The table below compares what each data source gives and what it cannot answer.

The decision rule is simple. If you track reach and brand awareness, look at impressions. If you track whether traffic and conversion actually happen, look at referral clicks. Basing a budget decision on impressions alone is like mistaking window-shopping traffic that never converts for success.

Step by step: building the measurement dashboard

The setup has three layers: Search Console impressions, GA4 referral clicks, and a visualization that joins the two. The sequence runs like this.

1. Establish the Search Console impression baseline

Open the generative AI report, set the Date dimension to weekly, and use the Pages dimension to note which URLs are listed on the AI surface. If you have no access, leave this layer empty and proceed with GA4 alone.

2. Isolate the AI referrers in GA4

In the Traffic acquisition report, select the session source/medium dimension and filter the chatgpt.com, perplexity.ai and, where present, chat.openai.com sources. Collecting them into one "AI referrers" channel group keeps the comparison sustainable.

3. Merge both layers on one calendar

In a tool like Looker Studio, plot Search Console impressions and GA4 clicks as two lines on the same date axis. The goal is not to ratio them but to see trend alignment: does a rise in impressions get followed by a rise in clicks, or does visibility fail to convert? For choosing the right tool, our AEO tools list gives a starting point.

Common mistakes in measurement

The most expensive mistake is treating the two data sources as one metric and adding them. Search Console impressions and GA4 clicks come from different systems with different definitions; adding them arithmetically produces a meaningless number. The following are common traps that quietly break measurement.

• Mistaking impressions for traffic: High impressions are possible with nobody reaching your site. Do not declare success without reading the click side in GA4.
• Mistaking one engine for the whole universe: Search Console covers Google only. Perplexity or ChatGPT traffic is captured only by GA4 referral; they are separate channels.
• Mistaking the report's absence for a loss: Since the report is limited to a subset, not seeing it does not mean you are invisible; in that case GA4 becomes your only source.
• Expecting queries: There is no query dimension in the generative AI report; you have to make content decisions from page, country and device.

If you want to deepen the general framework of measurement logic, the guide on what AEO is builds the concepts of visibility and answer engines from the ground up.

Our own measurement: a 12-question visibility test

Once the dashboard is running, the question becomes: what is the number? On 2026-08-29 we ran our own measurement for ogocer.com. We asked a search-grounded model 12 questions and counted the sources each answer cited.

Result: we were cited in 1/12 queries, or 8.3%. The single query where we appeared placed us first among 8 cited sources, alongside cryptonews.com, binance.tr, apple.com. We were absent from the other 11 queries. Across all answers, 139 unique sources were cited.

The most telling row is youtube.com. A video source appearing this often in answers to text-based questions shows visibility is not built by blog posts alone.

A single measurement means little. The value is in repetition: run the same 12 questions again and the difference becomes first-hand evidence of what your work changed. Write down the method, stamp the date, keep the query set fixed.

What to do after you measure visibility

Measurement is a diagnostic tool; the real work is closing the gap the number reveals. If impressions are low, you look at content and citation earning; if clicks are low, you look at title and snippet clarity. After gathering the figures, three scenarios emerge.

• Impressions yes, clicks no: AI lists you but the user does not click. Work on the content's citation value and brand clarity.
• Impressions no, clicks no: You are not entering answers at all. Visibility tactics take over.
• Impressions no but clicks yes: You are probably arriving from non-Google engines; break down the channel mix in GA4.

Once you have built your dashboard and gathered the first two weeks of trend, if you want to clarify which scenario you are in, we can plan the visibility strategy together. To start, you can reach us through the AEO services page. The number comes first; strategy is built on top of that number.