After nearly two years of AI search anxiety, Google has finally given publishers first-party visibility into generative AI features inside Search Console. The catch: the report is built on impressions, not clicks. You can see where your pages surface in AI Overviews or AI Mode, but not how much traffic or revenue is at stake.
The counting rules make that limitation concrete. As John Mueller explained, impressions are based on “links to your site being shown in AI Overviews / AI Mode.” A directly visible link counts immediately, while a link hidden behind an activation only counts after someone expands it. That is a presence metric, not an outcome metric.
Visibility is not exposure
For performance marketers, agency leads and D2C founders, raw AI impressions can be dangerously seductive. A page can pick up thousands of AI impressions without generating a single qualified visitor or sale. To make the report useful, layer it with commercial data and traffic dependency.
- Search Console’s Generative AI report for subfolder- and page-level AI impressions
- GA4 or your analytics platform for traffic mix, direct share and returning audience
- One primary commercial KPI, such as revenue, conversions or leads
- Optional AI Overview prevalence and server-log crawler activity for extra signal
Start by exporting the report at a subfolder level, such as /news/, /sport/ or /business/. Keep in mind the pages export has a 1,000-row limit, so once you exceed it you are working with a sample.
Build a commercial exposure model
The core approach is a normalized 0-100 score that combines AI visibility, Traffic Dependency and Commercial Value. For most teams, the formula is Core AI Commercial Exposure = (V × T × C)^(1/3). If you have reliable query-level AI Overview saturation, add Substitution to get (V × S × T × C)^(1/4). Do not enter zero when you are missing AI Overview data; zero means you measured substitution and found none.
This changes prioritization. One section might show high AI visibility but also high revenue, while another has comparable visibility and far less commercial value. A third could have low visibility but a large share of revenue. The score highlights where business impact is likely concentrated, which is far more useful than a visibility trend line.
Then measure resilience and opportunity
Exposure is not the same as risk. Two sections can have identical exposure scores but very different protection. Score resilience by averaging branded search, direct audience, returning audience and a content defensibility score. The higher the score, the better a section can withstand AI-driven substitution.
Server logs add a layer no dashboard gives you. Segment AI crawler activity into training bots, search bots and retrieval bots, then map frequency and breadth by section. High-value proprietary content being hammered by retrieval crawlers may signal a licensing or partnership opportunity, not just a threat.
The practical move is to stop waiting for Google to add click data. Build your own model, attach visibility to money and dependency, then use resilience and crawler demand to decide where to invest, block or negotiate.
Source: Search Engine Journal
