Can You Actually Measure Whether Your Website Appears in Google’s AI Search?

For the last few years, businesses have been told they need to prepare for AI search.

The problem was that measuring it was difficult.

You could see an AI Overview in Google, notice your website being referenced, or use a third-party visibility tool. But there was a large gap between seeing something happen and having reliable first-party data showing how often it happened.

That is beginning to change.

In June 2026, Google introduced dedicated Generative AI performance reports in Search Console, covering visibility in AI Overviews and AI Mode. The reports are still being rolled out, but for businesses that have access, Google can now show when pages from their website appear within its generative AI search features. Google explains the Generative AI performance reports here.

So, can you now measure whether your website appears in Google’s AI search?

Yes — but only to a point.

The new data makes AI search visibility much less mysterious. It does not, however, give businesses a complete picture of why Google selected a page, what every user asked, or some new universal “AI ranking” score.

That distinction matters.

What Google Search Console can now show you

Google’s Generative AI performance report records impressions from supported generative search features.

At the time of writing, that includes:

• AI Overviews;
• AI Mode.

Google also provides a separate generative AI report for Discover.

Within the Search report, businesses can break their visibility down by:

• page;
• country;
• device;
• date.

Google says the report can therefore show which URLs have appeared in its AI features, where the searches originated, which devices were used and how visibility has changed over time. Google’s announcement provides more detail on the reporting.

This is much more useful than simply knowing that a page appeared in an AI Overview once.

If a particular article repeatedly generates AI impressions over several months, for example, you can begin to see that Google is regularly finding it relevant within generative results.

Equally, if most of your AI visibility comes from three pages rather than 300, that is useful information about the type of content Google is surfacing from your site.

An AI impression doesn’t mean quite the same thing as a traditional ranking

This is where businesses need to be careful with the numbers.

Traditional SEO reporting has accustomed us to concepts such as rankings, clicks, impressions, click-through rates and average positions.

Generative search works differently.

Google describes an impression in the AI report as a link to your site being shown to a user within a generative AI feature. The exact way impressions are counted depends on the Search feature involved.

For example, content inside an AI Overview may need to be brought into view by the user before the associated link receives an impression. AI Mode can also involve follow-up questions, with each follow-up effectively creating another search interaction and another opportunity for links to appear. Google documents how Search Console impressions and clicks are counted.

That makes “we received 5,000 AI impressions” useful data, but not necessarily useful in isolation.

It doesn’t automatically tell you that you were the most prominent source, that someone read your content, or that those impressions generated business.

The same principle already applies to ordinary Search Console reporting, but it becomes even more important when the result itself may contain an AI-generated answer assembled from multiple sources.

What Search Console still doesn’t tell you

The new report closes an important measurement gap. It doesn’t close every one.

Search Console does not suddenly reveal:

• why Google’s systems selected a particular page;
• Google’s internal AI or ranking signals;
• a universal ranking position inside AI results;
• every prompt or question associated with an appearance;
• which particular fact or passage Google considered useful;
• complete user behaviour after somebody saw the generative result.

Google also excludes experiments taking place within Search Labs from Search Console reporting while those experiments remain under active development. Google’s documentation explains the limitations of generative AI reporting.

There is therefore still a sizeable difference between measuring visibility and completely understanding generative search.

That is worth remembering when evaluating software or services promising precise “AI rankings”.

Google itself now warns website owners to be cautious of third-party tools claiming to use internal Google metrics. Third parties do not have access to Google’s internal ranking or AI systems. Google addresses this in its guidance on AI features and Search.

That does not make third-party AI visibility tools useless. They can still monitor selected prompts, compare appearances or help identify patterns.

But their results should be understood for what they are: their own measurements of observable outputs, not direct access to how Google’s systems work.

Does this mean businesses now need GEO instead of SEO?

Probably not — at least not for Google.

Terms such as GEO, or Generative Engine Optimisation, and AEO, or Answer Engine Optimisation, have become increasingly common as AI-generated search has expanded.

There are genuine changes in how people search and how search results are presented. Businesses should pay attention to those changes.

But that doesn’t automatically mean Google Search now requires a completely separate optimisation discipline.

Google’s own 2026 guidance is unusually clear on this point.

The company says that the foundations of visibility in generative AI Search remain normal SEO principles: making content accessible to Google, maintaining a clear technical structure and creating useful, distinctive content for people. See Google’s guidance for succeeding in AI Search features.

Google also directly addresses several popular AI optimisation theories.

For Google Search, it says:

• you do not need a special llms.txt file;
• you do not need special AI-specific markup;
• you do not need to split every article into small “chunks” for an AI system;
• you do not need to rewrite content in a special AI-friendly style;
• there is no special schema markup required for generative Search;
• you do not need to manufacture mentions around the web simply to influence AI systems.

Structured data can still be valuable for normal SEO purposes. Clear headings still help readers. Well-structured information is still useful.

The distinction is that these things shouldn’t be done because somebody has claimed they are secret requirements for appearing inside AI Overviews.

AI search has changed since Google’s SGE experiments

When we previously wrote about Google’s Search Generative Experience and SEO, generative search was still largely something businesses were trying to prepare for.

The discussion was understandably speculative.

What happens to organic traffic if Google answers more questions directly? How should websites adapt? Would generative results fundamentally change SEO?

Those questions haven’t disappeared, but the conversation has moved on.

Google now has AI Overviews and AI Mode as identifiable Search experiences, and Search Console can provide first-party reporting for generative visibility on eligible websites.

We can therefore start replacing at least some speculation with actual data.

That doesn’t mean every uncertainty has been resolved. It means businesses now have a better way of separating “AI search seems important” from “these pages are actually appearing in Google’s AI results.”

What should you look for in the Generative AI report?

If your Search Console property has access, we wouldn’t recommend treating the total number of AI impressions as another headline KPI to chase.

Start with the pages.

Which pages are actually appearing?

Look at which URLs Google is surfacing.

Are they:

• service pages;
• detailed articles;
• research;
• product documentation;
• comparison content;
• FAQs;
• something unexpected?

The answer can tell you considerably more than the overall impression total.

A website may discover, for instance, that most of its generative visibility comes from a small group of genuinely detailed resources while generic pages barely appear at all.

That is a useful editorial signal.

It isn’t proof of a particular ranking factor, but it tells you which existing content Google is actually using within these experiences.

Is visibility increasing or decreasing over time?

A trend is usually more useful than a single day’s result.

Look at the data over weeks and months.

If AI visibility grows after you publish a genuinely useful collection of resources, that is interesting.

If it falls, investigate before jumping to conclusions.

Search Console itself can experience reporting errors. Google recorded one affecting Generative AI Search impressions between 13 and 17 August 2026, which caused reported impressions to decrease even though the issue related to data logging rather than actual Search visibility. Google records Search Console reporting issues here.

That is a good reminder that dashboards measure systems; they aren’t the systems themselves.

What is different about the content that appears?

This is probably the question we’d spend the most time investigating.

If five pages receive substantial AI visibility and another 50 don’t, compare them.

Do the visible pages contain:

• original expertise;
• detailed explanations;
• first-hand observations;
• proprietary data;
• clear comparisons;
• useful examples;
• information that isn’t simply repeated from dozens of other websites?

Google’s own guidance increasingly emphasises what it calls non-commodity content: material that contributes something beyond information already widely available elsewhere. Google discusses distinctive and useful content in its AI Search guidance.

That is a more useful editorial question than asking how many times the phrase you’re targeting appears on the page.

Don’t optimise for the report itself

There is a danger whenever a new marketing metric appears.

People start optimising the metric rather than the underlying outcome.

We’ve seen the same thing happen with rankings, impressions, engagement rates, cost per click and countless other measurements.

AI impressions could easily become another one.

A company might celebrate doubling its generative AI impressions while receiving no useful enquiries, sales or meaningful engagement from the content involved.

Alternatively, an article that rarely generates a direct click might still introduce a brand to thousands of people while Google answers their question.

Understanding that relationship is going to take more than watching one number.

The Generative AI report should therefore sit alongside your wider Search Console, analytics and commercial data rather than replacing them.

For most businesses, the useful question is not:

“How many AI impressions did we get?”

It is:

“Is our useful content becoming visible when people use these new ways of searching, and does that visibility contribute to something valuable?”

You probably don’t need an entirely new content strategy

The arrival of measurable generative visibility doesn’t mean businesses should abandon everything they know about creating useful websites.

If anything, Google’s current guidance points in the opposite direction.

Technical SEO still matters because Google needs to discover and understand your pages.

Useful content still matters because Search needs something worth retrieving.

Original expertise matters because generic information is increasingly easy to reproduce.

A clear website still matters because users may click through after encountering your business within an AI result.

There will undoubtedly be new techniques, technologies and forms of search worth investigating as AI develops. It would be equally wrong to pretend nothing is changing.

But businesses should distinguish genuine changes from a new vocabulary being applied to familiar principles.

For Google Search today, the sensible approach is to measure the AI visibility Google actually gives you, understand which content is earning it, and keep improving the underlying website and information rather than chasing supposed GEO shortcuts.

Google’s AI search is becoming more measurable.

That makes it easier to make better decisions.

It doesn’t make it less important to ask whether the thing you’re measuring actually matters.