Your site ranks first on Google for certain keywords, but visits from those searches have collapsed — and you don't know why. The answer is probably in the data Google is already giving you for free, and that almost no one knows how to read correctly. Understanding that signal tells you whether you're losing traffic because of the AI spaces Google has placed above the results, and what you can do to get back into the citation instead of staying out of it.
Open Search Console, go to Performance, isolate the long-tail informational queries and look at their CTR. Today GSC still doesn’t have an official “AI Overview” segment — the filtering has to be done by intent, by hand, cross-referencing average position and CTR. When you see CTR collapse at the same position, you’re inside the new standard. And Search Console gives you the numbers for free, while everyone else is selling dashboards at 200 euros a month to tell you the same thing.
Throughout the articles in this series I’m showing you how to measure your visibility in AI answers. Today we’re talking about the most underrated tool: the one you already have installed, that you’ve used for years, and that a few months ago started giving you precise signals about what happens when Google places an AI Overview above your organic results.
What Search Console actually measures when an AI Overview appears
When Google shows an AI Overview, your page can end up in three states: cited as a source inside the AI box, positioned below the box in the classic blue results, or both things together. The figure that Search Console returns is the sum of impressions and clicks on that query, regardless of which “layer” of the SERP generated the interaction.
This is the crucial point — and it has to be said honestly, because here we’re entering the territory of operational deduction, not documented fact. Google hasn’t released a paper formally explaining how the AI Overview is counted within Search Console. What we see working with real client data is a pattern: informational queries where an AI Overview appears show a drop in CTR at the same average position. From this it follows that, even without an explicit “AI Overview impression” metric, the signal is there — you just have to know how to read it.
Why this data sits upstream of any paid tool
In previous articles I talked to you about how AI engines think and about E-E-A-T for AI. All that upstream work — building authority, structuring content, getting your entity recognized — serves one single purpose: appearing in generative answers. Search Console is the first place where this visibility becomes a number.
The paid tools that today sell “AI Overview monitoring” are doing, in 90% of cases, something you could do yourself: scraping the SERP for selected queries, comparing it with GSC data to estimate the impact. That’s perfectly fine if you have the budget and high volumes. But if you run a jewelry shop in Vicenza or a coffee roastery in Naples, the honest starting point is Search Console.
The average hides the fact that informational queries are collapsing while brand-related searches make up for it.
The test you can run in 15 minutes
The operational flow is this:
- Open Search Console and go to Performance > Search results
- Set the time range to “Last 6 months” and compare it with “Previous period”
- Sort the queries by descending impressions
- Filter only the informational queries (who/what/how/why/when, or “best”, “difference”, “guide”)
- Look at the CTR column: if for the top 50 informational queries you see an average drop greater than 30% at the same average position, you’re in the AI Overview scenario
The binary threshold I use with clients is simple. CTR dropped more than 30% on informational queries with a stable average position = AI Overview active on those queries. CTR stable or grown = your content is probably being cited as a source inside the AI box, and that’s good news.
Create a sheet with three columns: query, CTR October 2025, CTR current month
What I’ve seen across five clients over the last six months
Indicative test, not a study. I followed five clients longitudinally — an artisanal wine e-commerce based in Verona, a notary office in Lecce, a maker of mechanical tooling in Bergamo, a boutique hotel in Siena and a veterinary practice in Pescara — comparing their Search Console data from October 2025 with that from March 2026. Small sample, but a pattern clear enough to be worth telling.
Informational queries like “difference between Amarone and Valpolicella”, “what you need to open a VAT number”, “how to choose lathe tools”, “what to see in Siena in two days” lost on average 38% of CTR at the same average position. Transactional queries like “buy Amarone online” or “notarial deed quote” stayed stable. For one of the five clients — the wine e-commerce, which had editorial content structured with FAQs and an inverted pyramid — organic CTR dropped, but overall traffic stayed unchanged because the page was cited as a source in the AI Overview and generated qualified clicks.
The honest limit: five clients are not a study, and today Search Console still doesn’t have a dedicated “AI Overview” segment — reading the phenomenon remains indirect, by inference on CTR at the same position. Real analysis requires professional tools and an observation period of at least nine months on a larger sample.
The mistakes I see most often when clients show me their GSC reports
Four recurring patterns, in order of severity.
The first: looking only at total impressions and clicks. If the total is stable, the business owner breathes a sigh of relief. Wrong. The average hides the fact that informational queries are collapsing while brand-related searches make up for it. You have to segment by query type.
The second: not comparing CTR at the same average position. If the position changes, the CTR drop may not depend on the AI Overview. You should always filter the queries where the average position has stayed within the ±0.5 range.
The third: ignoring the “Search Appearance” tab. Google hasn’t yet announced an official segment dedicated to AI Overviews, but it’s in that tab that new result types (FAQ, How-To, Video) have historically appeared. Checking it monthly makes you the first in your sector to see the data when — and if — it arrives.
The fourth: measuring only Google. Search Console tells you nothing about ChatGPT, Perplexity, Claude or the Gemini app. For those you need direct tests on the AI engines — which I told you about in other articles in the series.
What to do concretely starting Monday
- Open Search Console and put a fixed monthly check in your calendar (15 minutes)
- Create a sheet with three columns: query, CTR October 2025, CTR current month
- Enter your 30 main informational queries and monitor the drift
- When a query loses more than 30% of CTR, go and check the AI Overview manually on Google: see whether it cites you as a source
- If it doesn’t cite you, work on author entity recognition and on the content structure
Let me also tell you what not to do: don’t immediately buy an AI tracking dashboard at 200 euros a month. First exhaust what Google gives you for free. Once you’ve figured out which are your 30 critical queries, then you’ll assess whether a professional tool gives you a return on investment.
Where this series is going
Search Console is the first step in your system for measuring visibility in AI answers. In the next articles in the series I’ll show you how to cross-reference this data with direct tests on ChatGPT and Perplexity, how to build a simple dashboard in Looker Studio, and how to tell the difference between AI Overview cannibalization and classic organic ranking loss — two things that at a glance look identical but require opposite interventions.
The thread is always the same: measure first, act later. Google is already telling you a lot, for free. The point is to listen to it.
