Google's AI Visibility Report Gives You Impressions, Not Answers. Here's How to Build the Rest.

Google rolled out its Generative AI performance report to all websites worldwide as of August 31, 2026. Every demand gen leader with a Search Console login now has a new tab full of impressions data. Most of them are going to misread it.

The report tells you how often links to your site appeared in AI Overviews and AI Mode. It does not tell you clicks, CTR, average position, or which queries triggered those appearances. Harry Clarkson-Bennett, SEO Director at the Telegraph, frames it as "useful but incomplete": impressions alone don't quantify exposure, risk, or commercial value. The gap between what the report shows and what your CFO will ask about is where the real work starts.

The Measurement Trap You're Walking Into

Leadership sees "AI impressions" ticking up and asks: "Are we winning or losing in AI search?" Impressions can't tell you. They're a visibility signal, not a traffic metric, and definitely not a pipeline metric. Treating them as a KPI is a category error.

In a study of 68,879 Google searches, users clicked a traditional result on 8% of visits with an AI summary versus 15% without one. Users clicked a link within the AI summary itself on roughly 1% of visits where a summary appeared. Those are directional numbers from one dataset, not universal law. But they illustrate the core problem: AI visibility and AI-driven clicks are wildly different quantities, and this report only gives you the first.

Seer Interactive's panel across 53 brands and 5.47 million queries found that CTR on AI-Overview queries was about 1.3% in December 2025 and roughly 2.4% in February 2026, while queries without an AI Overview sat at around 3.8% CTR in February 2026. A single CTR headline doesn't generalize across your site, your ACV, or your funnel stage.

What to Actually Build With This Data

Clarkson-Bennett's framework is worth stealing and adapting for B2B demand gen. The core idea: layer AI impressions with traffic dependency, commercial value, and (optionally) AI Overview prevalence data to produce a composite exposure score.

Step 1: Export and classify. Pull page-level AI impressions from the Generative AI report. Classify URLs by content type: product pages, comparison pages, top-of-funnel glossary content, case studies. The 1,000-row export limit means you're looking at a sample for larger sites.

Step 2: Map commercial value. Pick one consistent commercial KPI per content area. For B2B SaaS, that's probably demo requests, qualified form fills, or pipeline dollars sourced from organic. Don't blend metrics across sections; the comparison breaks if /glossary uses page views while /pricing uses MQLs.

Step 3: Calculate traffic dependency. What percentage of each content area's sessions come from organic search? A glossary section pulling 85% of its traffic from Google is structurally more exposed than a product page where 60% is direct or branded. Most teams skip this variable, and it changes the risk picture dramatically.

Step 4: Build the exposure score. Clarkson-Bennett's formula: AI Commercial Exposure = (Visibility × Traffic Dependency × Commercial Value)^(1/3), with all inputs normalized to a 0-100 scale. High AI visibility on a section with low commercial value and low search dependency produces a low score. That's the point: it separates noise from genuine business risk. If you have third-party SERP data showing AI Overview prevalence for your key queries, fold that in as a fourth variable and take the fourth root instead. But don't enter a zero if you haven't measured it. Zero means "measured and found none," which is different from "didn't measure."

The Resilience Layer Most Teams Won't Build

Exposure tells you where you're vulnerable. Resilience tells you how vulnerable. Four inputs: branded search share, direct audience percentage, returning audience rate, and content defensibility (scored qualitatively, 0-100, based on proprietary data, expertise depth, and replication difficulty). Average the four. A section with strong branded demand and hard-to-replicate content is in a fundamentally different position than a glossary farm dependent on informational queries.

Scoring your own content defensibility honestly means admitting that some top-of-funnel pages are commodity content AI can reproduce without citing you. That's a finding, not a failure. It tells you where to invest and where to stop pretending organic will hold.

Connecting This to the Questions Leadership Actually Asks

Your CFO won't ask about AI impressions. They'll ask: "Are we losing organic pipeline to AI search?" The Generative AI report can't answer that directly. What you can build is a defensible narrative: here's where our AI visibility is concentrated, here's the commercial value at stake, here's our dependency on organic for that value, and here's our resilience score.

Success = a weekly dashboard that separates AI visibility trends from organic session trends from pipeline contribution, by content type. Guardrails = don't report AI impressions as a standalone KPI without the exposure and resilience context. Stop-loss = if you catch yourself building a slide that says "AI impressions up 40%" without the commercial exposure layer, stop. That slide will get you a budget question you can't answer.

The report doesn't integrate natively with GA4 or your CRM. Someone on your marketing ops team needs to own the export-classify-join workflow weekly, or it rots. That's real capacity cost. But the alternative (reporting raw impressions or ignoring the data entirely) leaves you defending a position with no evidence when the board asks about AI search impact.

Impressions are the floor, not the ceiling. The report Google shipped is a starting input, not a finished answer. The teams that build the exposure and resilience layers now will have six months of baseline data when the next algorithm shift hits. The ones waiting for Google to add click data to the report may be waiting a long time, for reasons Google would prefer not to discuss at dinner.