Google Generative AI Report Feels Broken — Here's How To Actually Make It Useful
Google finally gave SEOs first-party visibility into how often their content shows up inside AI Overviews and AI Mode. For a lot of teams, the excitement lasted about five minutes — right up until they realized the report shows impressions, but almost nothing else that’s actually actionable.
No clicks. No query-level data in most views. Just a rising or falling line that tells you your content is being seen somewhere inside Google’s AI features, with very little context on what that actually means for your business.
Here’s what’s changed, why the report looks this thin, and a practical way to turn it into something genuinely useful anyway.
Google's Generative AI Report Shows Visibility, Not Value
Search Console’s Generative AI Performance report tracks how often your pages appear in AI Overviews and AI Mode.
Key facts:
- The report gives sitewide and page-level impression data for content appearing in Google’s generative AI features.
- It does not include click data or reliable query-level detail for most sites.
- According to Google’s own John Mueller, impressions are counted based on links to a site being shown inside AI Overviews or AI Mode — a link shown directly can count without any user action, while a link that requires activation (like expanding a section) only counts once a user actually activates it.
- The page-level export is also limited to roughly the first thousand rows, so larger sites are typically seeing a sample rather than the complete picture.
Why This Matters
Impressions alone answer one question — “is my content showing up in AI features?” — and leave the far more important ones unanswered: is that visibility costing us clicks, and does it matter for the business either way? Without click or query data, it’s genuinely hard to separate a section of your site that’s thriving inside AI answers from one that’s quietly losing traffic to them. That gap is exactly why treating this report as a standalone dashboard, rather than one input among several, leads a lot of teams to write it off entirely — when the smarter move is to combine it with data you already have.
Turning Visibility Into A Real Exposure Score
The fix isn’t waiting for Google to add more columns. It’s layering the AI visibility data against information you already collect elsewhere.
Key facts:
- A workable exposure model combines four inputs: AI Visibility (where impressions concentrate by subfolder or content type), Traffic Dependency (what share of a section’s traffic comes from organic search), Commercial Value (revenue, leads, subscriptions, or conversions tied to that content), and, optionally, AI Substitution (how often relevant queries trigger an AI Overview at all, using third-party SERP data).
- A simple version of the formula — without the optional substitution step — multiplies AI Visibility, Traffic Dependency, and Commercial Value together and takes the cube root, producing a single 0–100 score per content area.
- The logic is deliberately conservative: heavy AI visibility on a page with almost no commercial value shouldn’t produce a high exposure score, and a highly valuable page with little AI visibility or search dependence shouldn’t either.
Why This Matters
This reframes the conversation from “our AI impressions went up” to “here’s exactly which parts of the business are most exposed if AI search keeps growing.” That’s a very different conversation to bring into a leadership meeting, and it’s the kind of analysis that turns a thin report into a genuine planning tool — one built from data most SEO and analytics teams already have sitting in Search Console and their analytics platform.
Not All Exposure Is Equally Risky — Resilience Fills The Gap
Two content areas can carry the same exposure score and still face very different levels of real risk, depending on how defensible that content actually is.
Key facts:
- A resilience score can be built from four inputs: branded search volume (demand specifically for your brand name), direct audience share (traffic that comes straight to your site rather than through a search intermediary), returning audience share (how much of your traffic is repeat visitors), and content defensibility — a qualitative judgment of how hard your content is to replicate, based on its uniqueness, effort, expertise, and access to proprietary information.
- The first three inputs can be measured as straightforward percentages from your analytics platform. Content defensibility needs a manual, consistent scoring framework, since it’s a judgment call rather than a metric.
- Averaging the four gives a single resilience score, where a higher number means a content area is better protected against further substitution by AI-driven search.
Why This Matters
A site or section with a strong direct and returning audience, high brand-search demand, and genuinely hard-to-replicate content is in a fundamentally different position than one relying almost entirely on generic organic traffic for informational queries — even if both show identical AI visibility numbers. Resilience scoring is what separates “this looks exposed but is actually fine” from “this looks fine but is quietly vulnerable.”
Server Logs Add A Demand Signal Google Won't Give You
Key facts:
- Server logs can reveal AI crawler access patterns by bot type, letting teams separate training bots, search bots, and retrieval bots instead of lumping everything together as generic “AI traffic.”
- This creates a proxy for AI demand: what content is being accessed, how often, and by what kind of crawler.
- Crawler access alone doesn’t prove that content was used to train a model or that any AI company intends to pay for it — it only shows that the content is attracting AI crawler activity consistent with training, search, or retrieval behavior.
Why This Matters
Combined with commercial value and content defensibility, server-log demand data can point to a genuine opportunity rather than just a threat. High-value, hard-to-replicate content that’s being heavily accessed by retrieval crawlers is exactly the kind of asset worth a closer look before deciding whether to block AI crawlers, pursue a licensing conversation, or lean further into that content area — decisions that are far easier to justify with demand data behind them than with a gut feeling alone.
Theme Of The Week: Turn A Thin Report Into A Real Framework
Google’s Generative AI report, taken on its own, tells a limited story: content is showing up somewhere inside AI features, and that’s about it. But paired with the traffic and revenue data most teams already track, it becomes the foundation of a real framework — one that separates simple visibility from actual business exposure, and actual exposure from resilience.
The report was never going to hand teams a finished answer. The teams that get real value out of it are the ones willing to build the missing layer themselves, using data they already have.
Key Takeaways
- Google’s Generative AI report shows visibility, not value. Without click or reliable query data, impressions alone can’t tell you whether AI features are helping or hurting a given content area.
- Exposure comes from combining data sources. Layering AI visibility with traffic dependency and commercial value turns a thin report into a genuine, decision-ready exposure score.
- Resilience explains why identical exposure scores can mean very different risk. Branded search demand, direct and returning audiences, and content defensibility all affect how vulnerable a section really is.
- Server logs reveal opportunity as well as risk. AI crawler activity, layered against content value, can point toward licensing conversations or investment decisions — not just defensive ones.
Frequently Asked Questions
1. What does Google’s Generative AI report in Search Console actually show?
It shows sitewide and page-level impression data for content appearing in AI Overviews and AI Mode. It does not include click data, and query-level detail is limited or unavailable for most sites.
2. Why doesn’t the Generative AI report include clicks?
Google hasn’t fully explained the decision, but John Mueller has clarified that impressions are counted based on links being shown inside AI features — a directly shown link can count without any user action, while a link requiring activation only counts once activated.
3. How can a business turn AI visibility data into something actionable?
By combining AI visibility with existing traffic and commercial data — building a simple exposure score from AI visibility, traffic dependency on organic search, and commercial value like revenue or conversions.
4. What is AI resilience, and why does it matter?
AI resilience measures how well-protected a content area is against further substitution by AI search, based on branded search demand, direct and returning audience share, and how difficult the content is to replicate. Two areas can have the same exposure score but very different resilience.
5. Are server logs necessary to analyze AI visibility?
No, they’re optional. A solid exposure and resilience analysis can be built using just Search Console and analytics data. Server logs add an extra layer of detail by revealing which AI crawlers are accessing specific pages and how often.
6. Does AI crawler activity mean an AI company will license or pay for that content?
No. Crawler access only shows that content is attracting activity consistent with training, search, or retrieval behavior. It doesn’t prove the content was used for training, and it doesn’t guarantee any licensing or payment opportunity — it simply flags content worth evaluating further.
