What Google’s New AI Search Reporting Cannot Yet Measure

Google has begun testing a Search Generative AI Performance report in Google Search Console, currently limited to a subset of UK website owners. It tracks how often content appears in AI Overviews, AI Mode and AI Overviews in Discover, broken down by page, country, device and date. For the SEO community this is the first time AI-specific visibility data sits inside the tool teams already use to measure organic performance.

It is also a measured beginning rather than a complete answer. The report records impressions and nothing else, with no click data attributable to AI surfaces and no query-level information. For a discipline that has built its operating model around the relationship between impressions, click-through rate and downstream conversion, the absence of CTR is the more consequential gap. The question it raises, and the one we think is most worth focusing on, is what happens when SEO measurement loses the metric that has historically bridged visibility and value, and what should fill that space as the reporting matures.

The CTR Gap and Why It Matters

For more than a decade, SEO reporting has rested on a clear sequence in which impressions show how often a page appeared, click-through rate shows how often those impressions converted to visits, and sessions, conversions and revenue follow from there. The SEO reports a B2B agency produces every month are built on that chain, with the Search Console Performance report providing the underlying data.

AI search disrupts the sequence at the second step. AI Overviews and AI Mode often answer a user’s question in place, with citations to source pages presented alongside or beneath the synthesised response, which means whether a user reads a citation and whether they click through to it are now distinct events, and only one of them is reflected anywhere in current measurement.

The new Search Console report covers the first half of that sequence for AI surfaces, confirming that a page has appeared in front of users via AI without recording what happened next. That is not so much a shortcoming of the reporting as a reflection of the harder underlying problem of measuring outcomes from generative search. The practical effect for SEO teams is that they are reading one side of a balance sheet, and the open question is how long the other side stays missing.

What Impressions Alone Can and Cannot Show

Impressions on their own carry specific weight, showing presence and trajectory. A page appearing in AI Overviews more often this month than last is being read by the AI layer more often, which suggests the relevance and authority signals around it are working, while a page that has stopped appearing is being deprioritised by the AI for some reason. Both observations are useful directional signals, which is why we would not dismiss the new report as incomplete despite the gaps.

What impressions cannot do, on their own, is answer the question stakeholders ultimately care about, which is whether AI exposure is contributing to commercial outcomes. Without click data, session data or conversion attribution from AI sources, the connection between visibility and value is left to be inferred rather than measured.

Interim proxy approaches help, even if they only go so far. Branded search lift studies, comparing branded search volumes before and after sustained AI presence, can suggest whether AI exposure is driving subsequent return visits through other channels, and direct traffic anomalies during periods of high AI impression growth may indicate that users are arriving via memory rather than a tracked citation click. Neither approach is precise, but both are better than treating AI impressions as a standalone success metric, which is the temptation a stakeholder unfamiliar with the limitation will reach for first.

The pragmatic interpretation for now, in our view, is to report AI impressions as a separately labelled visibility metric clearly distinguished from organic performance data, and to set stakeholder expectations openly about what it does and does not measure.

Impressions confirm that pages are being read by the AI layer. They do not confirm that the exposure is delivering commercial value. The harder question, how often AI appearances translate into outcomes, remains unanswered until Google adds the underlying data.

The Metrics and Terminology Likely to Emerge

Google has been explicit that additional metrics will be introduced over time, and the industry should expect the vocabulary of AI search measurement to expand significantly over the next twelve to eighteen months. Several candidate metrics are plausible based on how analogous problems have been solved before, and a few of them are likely to become standard reporting fixtures within a year.

• Citation click-through rate. The rate at which users click on a citation within an AI response. This is the most direct equivalent to traditional CTR and, in our view, the single most useful addition Google could make to the report.
• AI citation share. The share of relevant queries in which a site appears among the AI’s cited sources, conceptually similar to share of voice in paid media. This is the most likely candidate for a standard agency reporting metric, because it translates cleanly into a competitive position the way ranking position once did.
• Citation position and weighting. Where in an AI answer a citation appears, since visibility of the first citation differs meaningfully from the third or fourth.
• Synthesis weight. Whether a citation contributes to the synthesised text of an AI response or sits below it. The two carry different value signals and are likely to be reported separately once Google has the data to do so.
• AI brand lift indicators. Aggregated signals showing whether AI exposure correlates with later branded search or direct visits, drawing the line back to commercial outcomes that impressions on their own cannot.
• Outcome attribution from AI sources. The hardest of these to deliver, since it requires some form of identifier passed through citation clicks, and accordingly the slowest to arrive.

Terminology tends to follow the metrics, and as Google introduces named features in the reporting interface, industry shorthand catches up. The disciplines of generative engine optimisation and answer engine optimisation, which until recently operated without formal measurement tools, are starting to acquire them, and SEO teams should expect to retrain their stakeholder vocabulary alongside their own.

The other change worth anticipating is to the structure of monthly reporting itself. The current Performance report presents a unified view across all Search appearances, but as AI metrics mature, the most likely outcome is a tiered structure that separates standard Search, AI surfaces and any other appearance types into distinct sections, each with its own headline metrics. Agencies that build that structure into their reporting templates now will be better placed when the data arrives to populate it.

How SEO Reporting Should Adapt Now

For SEO leads managing UK B2B sites with access to the test, three steps are worth taking immediately.

1. Add a dedicated AI visibility section to monthly and quarterly reporting, distinct from standard organic data and labelled clearly as impressions-only so it cannot be misread as a traffic or conversion measure.
2. Build the interpretation in front of the data rather than behind it. A short note explaining what AI impressions measure, what they do not measure and how the team is reading the directional signal is more useful to stakeholders than the data on its own.
3. Choose at least one proxy outcome signal to track alongside AI impressions for the period before Google’s reporting expands. Branded search trends, direct traffic by landing page or a custom GA4 segment tied to high-AI-impression pages are all defensible options.

For teams whose markets sit outside the UK rollout, the priority is to prepare reporting templates now so AI data can be slotted in seamlessly when the global release arrives. Bing has been offering comparable AI performance reporting globally for some time, and teams running cross-engine measurement should accommodate both data sources in their reporting architecture rather than treating Google as the only relevant source.

The Implication for UK B2B Sites

The UK-first rollout has an effect that is easy to underestimate, since UK-based brands will become familiar with AI visibility data, and the gaps in it, ahead of competitors in larger markets. That familiarity will shape how teams interpret the global release when it lands, and how the reporting templates the rest of the industry adopts get designed.

Alongside the measurement question sits a substantive technical point. AI surfaces favour content with clear structure, demonstrable expertise and the trust signals that have always supported organic ranking, all of which trace back to Google’s E-E-A-T framework and the integration of helpful content signals into core ranking systems. Investment in technical SEO, content depth and structured data continues to be the most reliable route to AI citation, regardless of how the surrounding measurement metrics evolve.

The new report is the beginning of AI search measurement rather than the end of it. Impressions are the first metric available, click and query data should follow, and the broader vocabulary will keep evolving for at least the next year. The SEO teams that lead this period will be the ones who treat the gaps in current reporting as a planning prompt rather than a reason to wait.

Priority Pixels works with B2B businesses across AI SEO, technical SEO and content strategy to build measurement frameworks that reflect how search is changing. The new Search Console reporting is one input among several, and its value depends on the rigour of the reporting context built around it, and on the willingness to track the right proxies while waiting for the metrics that will eventually arrive.