Google Can Now Show You AI Visibility. It Still Can’t Show You What It Caused.

Google AI visibility just became considerably easier to see.

As of August 31, Google has rolled out its new Generative AI performance reports in Search Console to websites worldwide.

Site owners can now see how often their URLs appear in generative AI features across Search and Discover, including AI Overviews and AI Mode.

You can see which pages appeared, where those appearances occurred, which devices were involved in Search, and how visibility changes over time.

That is meaningful progress.

But it is important to understand what Google has given us.

We now have better evidence of AI visibility. We do not have complete attribution for what that visibility caused.

Those are two very different things.

What Google’s New AI Reporting Actually Shows

Google’s new reports provide dedicated views into generative AI performance across Search and Discover.

They currently include:

impressions
pages
countries
devices for Search
dates, with hourly through monthly views

The generative AI data still contributes to Google’s overall Search and Discover performance totals. The difference is that we can now examine AI visibility separately rather than trying to infer all of it from broader Search Console data.

That matters.

For the first time inside Google’s own reporting environment, a publisher can look at a page and say, in effect:

Google is using this asset across its generative experiences.

But there are also conspicuous things we cannot see.

The announced fields do not include the queries that produced those appearances. Nor does Google currently list clicks as a metric within the dedicated generative AI reports.

So we can see that something happened.

We cannot reconstruct the entire path through which it happened.

That is where I think we need a more precise distinction.

Visibility Evidence Is Not Visibility Attribution

Visibility evidence tells us that our work is participating in discovery.

It might show that 75 pages from a website appeared across Google’s generative features this month. It might show that appearances are increasing, that older pages are continuing to surface, or that a growing portion of a knowledge estate is becoming retrievable.

Visibility attribution tries to connect those appearances to specific pathways and outcomes.

Which question caused this page to appear?

What did the AI response say?

Did the user notice our contribution?

Did that exposure influence a later branded search?

And did someone encounter the idea again elsewhere?

Did it contribute to trust?

Did it eventually lead to a visit, subscription, inquiry, or purchase?

Those are attribution questions.

And Google’s new reporting does not answer most of them.

That doesn’t make the visibility evidence less valuable.

It means we need to stop asking one kind of measurement to do the job of another.

AI Discovery Makes Attribution Harder

Traditional digital marketing trained us to like neat paths.

Search query → search result → click → landing page → conversion.

Even that journey was never as tidy as our dashboards sometimes suggested, but at least many of its steps were observable.

AI-mediated discovery introduces another kind of encounter.

Someone asks a question.

An AI system retrieves information from several sources, synthesizes an answer, and perhaps cites some of them.

The user learns something.

Maybe they click.

Maybe they don’t.

Or maybe they remember a company name or concept.

Maybe they encounter it again three weeks later.

Maybe the AI-generated answer resolves the immediate question while simultaneously increasing familiarity with the source behind it.

Influence can occur without producing a clean referral trail.

That creates a measurement problem, but not necessarily a value problem.

Evidence Can Tell Us Whether the Knowledge Estate Is Working

This is where I think the new Google reporting becomes especially interesting for experts building substantial bodies of work.

I explore the importance of connected intellectual surface area in The Surface Area Advantage, including how a larger knowledge estate creates more pathways through which expertise can be discovered.

Now we have another form of evidence for observing that process.

Imagine you have 300 substantive pages.

Over time, Google’s generative AI reporting shows that 20 pages are appearing. Then 45. Then 90.

Perhaps those pages span multiple intellectual clusters. Older assets continue appearing alongside newer ones. The countries broaden. Generative impressions increase.

You still cannot say:

These 90 pages generated exactly $X in revenue.

But you can say something strategically significant:

An increasing portion of our knowledge estate is participating in AI-mediated discovery.

That is visibility evidence.

And for a discoverability strategy, it matters.

Measure the System at More Than One Level

I think one of the mistakes we make with emerging discovery channels is expecting a single metric to prove the entire chain of value.

But different measurements answer different questions.

An impression can tell us whether something appeared.

A citation can tell us whether something was attributed.

Traffic can tell us whether someone visited.

A conversion can tell us whether someone took a measurable action.

None, on its own, tells us the complete story of how expertise became known.

This suggests a more useful measurement principle:

Measure each layer for what it can actually tell you.

Use visibility data as evidence that your knowledge is being retrieved and surfaced.

Use traffic data to understand visits.

And use conversion data to understand measurable actions.

Use branded search, direct traffic, inquiries, customer conversations, and other signals to understand broader recognition where appropriate.

Do not discard an upstream signal because it cannot prove a downstream outcome.

The Missing Query Data Matters, Too

There is, however, a genuine strategic limitation in Google’s current reporting.

Without query-level information, we cannot see precisely why a particular page was retrieved.

That means we can observe surface area without fully observing intent.

To me, that distinction is fascinating.

If ten seemingly unrelated pages begin receiving generative impressions, there may be a conceptual relationship among them that Google understands even though the reporting does not reveal the questions creating that relationship.

The reporting shows us the footprint.

It does not yet show us the entire map.

Third-party AI visibility tools, manual prompt research, referral data, and other forms of observation may therefore remain useful alongside Google’s first-party reporting.

Google has not made the AI discovery ecosystem completely transparent.

It has made one important part of it more visible.

Better Evidence Should Change How We Think About AI Visibility

For years, one difficulty with AI search has been proving that your work is participating at all.

Google’s new reporting changes that.

Generative AI visibility is now sufficiently important that Google has created dedicated Search Console views for observing it.

That’s significant.

But perhaps the more important lesson is about measurement itself.

We should not confuse what we can observe with everything that occurred.

A knowledge estate can participate in discovery without producing an immediately attributable click.

An idea can influence someone without leaving a clean conversion path.

Recognition can accumulate across encounters no dashboard will successfully stitch together.

So yes, measure your Google AI visibility.

Watch which pages participate. Watch whether your intellectual surface area expands. Watch whether older assets continue working. Watch how the pattern changes over time.

Just be precise about what the numbers mean.

Visibility evidence tells you that your work is entering the discovery system.

Visibility attribution tries to tell you what happened because it did.

Google has just given us considerably more of the first.

We should not mistake the absence of the second for evidence that nothing happened.