Measuring AIO: How to Track AI Overview Visibility in Search Console

Measurement has long been the weakest link in AIO (AI Optimization). You could structure content so that AI answers would cite it, but you could not see the result: are we showing up in AI Overviews or not? Everyone was reading sample-based estimates from third-party tools. The "Generative AI features" report in Search Console closes that gap with Google's own data — though the window it opens is narrower than most people assume. This article covers the measurement side of AIO end to end.

What Exactly Does the Report Measure?

The new section reports how many times a URL from your site was shown to a user inside generative AI features. The definition is precise: an impression is counted when a link to a website is shown to a user within a generative AI feature.

You can slice the data by page, country, device and date. In other words, you can see which of your pages are being used as sources in AI answers and how that changes over time. That is a far more solid foundation than the sample-based estimates third-party tools have been offering.

The report is rolling out gradually. Google is giving it to a subset of sites first so it can test and collect feedback. If it has not appeared in your account yet, that is not a setup error — your turn simply has not come.

Three Surfaces, One Number

This is the most commonly misread part of the report. AI Overviews, AI Mode and Discover's generative AI features are not reported separately; they all sit inside the same number. When you see impressions rise, there is currently no way to tell whether that came from AI Overviews, AI Mode or Discover.

That is a serious constraint strategically, because user intent differs sharply across those three surfaces. In AI Mode the user is in deep research mode; in Discover they are in passive discovery mode. Interpreting impressions from both the same way will mislead you.

Why Don't the Totals Add Up?

The second source of confusion: the property-level total and the sum of the page-level rows do not match. This is not a bug — the two views aggregate differently.

If two different URLs from your site are cited in a single AI answer, that can count as one impression at property level and two at page level. Adding up page rows to verify the property total is therefore wasted effort.

Practical rule: never mix the two views into one metric, and never use one as validation for the other. Pick a single view for trend tracking and stay with it.

Why Is There No Click or CTR Data?

This is the most debated gap. Queries, clicks, click-through rate, average position, where the citation appeared in the answer, conversion data — none of it is there. Google says it is still working with site owners to determine which additional metrics would be useful, but it has made no public commitment to adding the missing dimensions.

There are two reasonable readings of that gap. The first is technical: a "click" on a link inside an AI answer behaves differently from a classic blue link. Users frequently read the answer and leave without clicking at all, so CTR loses much of its original meaning.

The second is strategic: impression data is relatively harmless for Google, whereas clicks and CTR would make the impact of AI Overviews on publisher traffic measurable. Publishing that picture clearly would open a debate Google does not currently want to have. We do not know which reading dominates — and both may be true at once.

Setting Up AI Visibility Tracking

Despite the missing metrics the report is useful — as long as you do not expect the wrong thing from it. Use it as a diagnostic lens rather than a performance scorecard: it shows you which topics Google is willing to surface your content for.

• Put AI visibility next to organic performance. Over the same date range, pages with high AI impressions but low organic clicks are likely ones where the AI answer already satisfied the question and the user had no reason to click. On those pages you need to deepen the content and create a reason worth clicking for.
• Group pages by category. Looking at individual URLs is noisy. Split them into groups — product, blog, guide, category — and measure which content type gets cited most on AI surfaces. That feeds your content plan directly.
• Track multi-week trends, not daily swings. AI surfaces are still changing fast; single-day drops usually come from Google-side experiments. Meaningful signal shows up over at least four weeks.
• Do not blend generative and classic impressions into one metric. A composite "total visibility" number buries two very different user behaviours under a single figure and makes your reporting misleading.

What You Can and Cannot Measure

Seeing the picture plainly is the fastest way to use the report correctly:

Where Does This Fit in AIO?

AIO rests on three legs: making content extractable by AI, building topical authority, and measuring the result. For the first two, long-established SEO practice still largely applies; we covered the detail in what AIO is and how to do it and AI Overview citation factors.

The third leg — measurement — has until now rested on guesswork. This report ties it to Google's own data for the first time. That moves AIO out of the "nice to have" bucket and turns it into a line item you can compare before and after.

The practical setup: date your content changes, take the four weeks before a change as your baseline, and compare against the four weeks after. Impressions alone are not a measure of success, but as a directional indicator they are reliable.

Does This Change SEO Strategy?

Short answer: it changes measurement, not strategy. Getting cited on AI surfaces still runs largely through classic SEO — an accessible URL, strong organic ranking, clear heading structure and verifiable information. We covered the evidence behind that in our article on AI Overview citation factors.

What is new is that something we previously only guessed at can now be verified with Google's own data. That moves AIO work out of the "nice to have" bucket and turns it into a measurable line item. For what AIO is and how it differs from AEO and GEO, see our article on what AIO is and how to do it.

One final caution: because the report is new and rolling out gradually, your data history is short. Rather than making big strategic calls before at least a quarter of data has accumulated, focus for now on establishing a baseline.