Google AI Overviews: how to measure click changes

AI Overviews can change the way a person uses Google before deciding to visit a website. To responsibly measure click decline or gain, compare equivalent pages and queries in Search Console, look at impressions, CTR, and position, and compare the change to the real SERP. There is no column that isolates each Overview, so the analysis needs to work with evidence, not assumptions.

Direct answer:Start with a group of URLs and queries where AI Overviews appear frequently, compare full before and after windows, segment by device and country, and record title, ranking, seasonality, and demand changes. A lower CTR with stable impressions may indicate fewer available clicks, but it does not alone prove that an Overview caused the drop.

For companies in Londrina and other Brazilian cities, the point is not to treat AI search as the enemy of SEO. It's about separating what changed in visibility, what changed in user intent and what changed in the page's ability to transform a visit into contact. This reading speaks toZero-click measurement in the age of AIand with a strategy ofGEO to appear in AI responses.

What AI Overviews change in traffic reading

An AI Overview brings together information from different sources directly on the results page and can show supporting links. In some searches, the user finds an initial explanation before opening any organic result; in others, the summary increases curiosity and leads to a more qualified click. The effect varies by query, industry, device, language, competition, and decision stage. Therefore, a general domain curve rarely explains what happened.

The Search Console Performance report records clicks, impressions, CTR, and position for results that lead to your property. This data is indispensable, but it has aggregation rules: a view by property is not the same as a view by URL, and anonymized queries or less relevant rows may not appear in the table. Use the page view to investigate the affected content, and keep the property view as context, not the final diagnosis.

The four metrics that need to be read together

On mobile, swipe the table sideways to see all columns.

The most useful standard is not an isolated metric. If impressions remain similar, position does not move significantly, and CTR drops on queries with Overview observed, there is a sign that deserves investigation. If impressions and position drop together, the hypothesis of loss of ranking, change in demand or competition is stronger. If clicks fall, but conversions and qualified contacts do not fall, the business may be filtering out visits that are not very useful instead of missing a real opportunity.

Set up a baseline before completing anything

Choose a window long enough to reduce noise, typically four to eight weeks, and compare with a previous period of equivalent duration. When there is strong seasonality, also compare with the same time the previous year. Exclude early days in Search Console and note campaign launches, site changes, migrations, price changes, news, and broad Google updates. A crash shortly after any of these events still needs testing, not automatic completion.

Create a simple spreadsheet with URL, query group, intent, device, country, clicks, impressions, CTR, position and SERP observation. The objective is to find patterns: informative guides may behave differently than service pages; cell phone searches may react differently; and branded terms deserve a separate group from generic terms. Branded and non-branded query filtering, when available on the property, helps prevent awareness growth from hiding a drop in organic acquisition.

How to Investigate Page, Query, and SERP

Start with the pages with the biggest difference in clicks. Apply an exact URL filter, compare the date ranges, and then open the queries that contributed the most to the variation. For each relevant query, check that the content still responds to the intent, that the canonical page is the same, that the title is clear, and that there is a major position change. Then perform a manual search in a clean environment, in the appropriate country and language, to document the presence and format of an AI Overview. Don't use a single screenshot as proof of a trend: the SERP is dynamic and personalized.

SERP observation is to create hypotheses, not to replace data. Record date, device, query, Overview presence, displayed fonts, and other space-taking features like ads, maps, videos, or dashboards. Then look for the same pattern across multiple queries across the cluster. If the crash only happens in one URL, first review the content and indexing. If it happens on many URLs that share intent and AI capability, the hypothesis gains strength.

Don't confuse correlation with causation

A visible Overview and a drop in CTR in the same month do not prove causality. Ranking changes, better competitors, lower demand, a rewritten snippet, an unattractive title, a technical issue, or an algorithm update can all produce the same signal. Google's guide to traffic drops recommends investigating the shape of the curve, the affected pages and queries, and events that occurred over the period. This discipline protects the team from hasty fixes, like rewriting the entire site for the sake of a small sample size.

It’s also important not to promise a “lost clicks for AI Overviews” count. Search Console measures the performance of links leading to the website; it does not deliver universal attribution of every missed impression or click to a specific SERP feature. An estimate can be useful for prioritizing work, as long as it is presented as a scenario and accompanied by the criteria used.

Measure impact on the business, not just the session

For a service company, the decisive indicator is not just organic visits. Integrate the clipping of affected pages with GA4 or the analytics tool, observing forms sent, clicks on WhatsApp, calls, quote requests, sales and assisted conversions. Classify source and content in CRM when possible. A reduction in clicks from very broad queries can coexist with growth in leads from specific searches; In this case, insisting on recovering the entire volume may worsen the quality of the acquisition.

Define a panel with three levels. At the discovery level, track impressions, viewability per query, and feature presence in the SERP. At the visit level, track clicks, CTR, engagement, and landing pages. At the outcome level, track qualified leads, quote rate, revenue, and assisted conversions. The connection between them does not need to be perfect to be useful: it shows where a drop in clicks actually turned into a commercial loss and where content continues to create demand.

Actions that make sense after analysis

If the page holds impressions and loses CTR, review the headline and description promise, direct response at the beginning, and coverage of intent. If you lose impressions and position, audit content, internal links, indexing and competition before touching cosmetic elements. If Overview cites better sources, study what kind of evidence, updating, authorship, and structure make those sources useful to the reader. The sustainable path is to improve utility, experience and clarity, not try to manipulate a generated response.

AI-powered search-ready content still needs to be good for people: explain the topic accurately, show authoritative sources, answer specific questions, keep pages crawlable, and make the next step easier. Onesolid on-page SEO foundationand clear service pages help Google understand the offering and help the visitor arriving via a support link decide whether to move forward.

Monthly checklist to track AI Overviews

1. Separate pages and queries by intent, brand, country, and device.
2. Compare entire periods and remove preliminary data.
3. List the biggest changes in clicks, impressions, CTR and position.
4. Document the SERP in a representative sample, without relying on a single search.
5. Record internal changes and external events that may explain the curve.
6. Compare traffic loss or gain with assisted leads, sales and conversions.
7. Prioritize one hypothesis at a time and measure the effect of the correction in the next cycle.

Sources for the analysis

• Google Search Console: how Performance report data is aggregated
• Google Search Console: tasks and comparisons in the Performance report
• Google Search Central: How to Investigate Traffic Drops
• Google Search Console: clicks, impressions and position

When a pattern is found, keep a dated record of the query set and the SERP observations, then repeat the same comparison in the next reporting period. This prevents a temporary interface change from becoming a permanent strategic conclusion and gives the team a reliable baseline for future decisions.