AI Overviews and Organic CTR: How to Measure the Impact on Your Site

There is no defensible universal percentage by which AI Overviews reduce your site’s organic click-through rate. Investigate your own pages and query cohorts, and distinguish AI visibility from clicks. A before/after traffic change can coincide with AI features without being caused by them.

In short

• No single percentage applies to every site; published studies report different effects for different query sets.
• Calculate CTR only from compatible clicks and impressions, and keep AI visibility separate from clicks.
• Compare stable cohorts, state what stayed uncontrolled, and report business outcomes beside traffic.
• Being cited in the overview is a different objective from ranking, and is worth tracking.

This guide supplies an analysis worksheet. It does not present a customer Search Console export or an invented Vercatus traffic result.

Start with the current reporting limits

Google’s June 2026 reporting announcement describes dedicated generative-AI impressions views for Search and Discover, with page, country, device where applicable, and date breakdowns. Its update states worldwide rollout by 31 August 2026. Check the current interface and report help before building an analysis around a metric.

The available AI-impression view is not a clean query-level AI Overview click dataset. Google AI Overviews and AI Mode appearances must not be treated as interchangeable with ordinary Web impressions. Use the ordinary Search performance data for its defined clicks and impressions; use the AI view to describe its defined visibility.

Vercatus’s Search Console integration reads the supported search-performance data. It does not claim to import every new report in Google’s interface.

Calculate CTR only from compatible counts

For one consistently defined report and cohort, CTR equals clicks divided by impressions. Sum the counts before calculating the ratio; do not average page percentages when their impression volumes differ.

Do not divide ordinary Web clicks by generative-AI impressions and call the result “AI Overview CTR.” Those counts refer to different reporting populations. If a report does not supply compatible clicks, the required CTR cannot be calculated from it.

Build a stable comparison cohort

Choose pages and query groups relevant to the business. Separate branded and non-branded queries, intent, country and device. Retain the same date granularity, filters and property across periods.

Record important changes: rankings, seasonality, campaigns, site migrations, stock availability, tracking changes and major content edits. An informational article and a branded product query have different click opportunities, so a blended sitewide average can hide the real pattern.

Worksheet field	Why it matters
Periods and filters	Makes the comparison repeatable
Page and query cohort	Prevents changing traffic mix from masquerading as an effect
Clicks and impressions	Supplies compatible CTR counts
Average position	Adds ranking context, with aggregation limits
AI impressions, separately	Describes observed AI visibility without inventing clicks
Sessions and qualified outcomes	Connects the change to business relevance
Concurrent changes	Makes alternative explanations visible

Compare changes with appropriate restraint

If informative pages lose clicks while branded pages remain stable, that is a useful pattern to investigate. It does not prove that AI Overviews caused the loss. A change in query demand or source selection can produce similar results.

Comparable unchanged pages can provide context, but they may differ in intent, competition and seasonality. Any comparison should state why its groups are plausibly comparable and what remains uncontrolled. Avoid calling a rough matched cohort a randomized experiment.

Inspect real results alongside the export

For a small panel of important questions, save dated search observations in the relevant market. Record whether an AI answer appeared and which sources were shown. Treat these as bounded observations, not a reconstruction of every impression in Search Console.

Look at the answer task. A result that fully answers a simple factual question can create different visit incentives from one that helps a buyer choose a detailed product guide. That interpretation should inform editorial work rather than a blanket rewrite of every page.

Report the business decision

State what changed, what you can measure and what action follows. A lost low-value visit and a lost qualified enquiry have different implications. If traffic declines while qualified outcomes remain stable, report both instead of choosing the more alarming metric.

What published studies report, and the limits

The numbers below are other organisations’ measurements on their own samples. They show the range of what has been observed. They are not a forecast for your pages.

Source	What it measured	Result as reported
Seer Interactive	3,119 queries across 42 client organisations, 25.1 million organic impressions, June 2024 to September 2025	Organic CTR on queries with an AI Overview fell from 1.76% to 0.61%
Ahrefs	300,000 keywords, Search Console data, December 2023 against December 2025	An AI Overview correlated with a 58% lower average CTR for the top-ranking page
Pew Research	Browsing behaviour	Users clicked a traditional result in 8% of visits with an AI summary against 15% without, and clicked a link inside the summary in 1%

All three are in the graded evidence as correlational and population-specific (AI Overview CTR impact, grade C). The same body of evidence reports that, for brands cited inside an overview, Seer measured higher organic CTR than for uncited brands on the same queries in Q3 2025 (citation benefit, grade C). Read that as a reason to track citation, not as a promise.

What Search Console can and cannot tell you

Search Console’s performance report gives clicks, impressions, CTR and average position, grouped by query, page, country, device, search appearance and date (Search Console as ground truth, grade A). It does not guarantee every row: it returns the top rows by clicks, some queries are withheld for privacy, and results grouped by page are aggregated to the canonical URL. Absence from an export therefore means not observed in that dataset, not zero impressions (completeness and attribution, grade A).

A step-by-step worksheet

1. Pick the cohort. Informational pages and branded pages behave differently; separate them.
2. Fix the filters. Same property, country, device, search appearance and date range in both periods.
3. Sum before dividing. Total clicks over total impressions for the cohort, never an average of page percentages.
4. Note the confounders. Rankings, seasonality, campaigns, migrations, tracking changes, major edits.
5. Add outcomes. Sessions and the qualified actions that matter to the business.
6. Save the observations. For a small panel of questions, note whether an AI answer appeared and which sources it showed.

A worked example, with invented numbers

A fictional guide site compares the same 40 informational articles across two periods. Clicks fell, impressions were flat, and branded product pages were stable. The tempting conclusion is “AI Overviews took our traffic”. The better statement is: “Clicks on informational queries fell by a stated amount while impressions held; branded pages did not move; AI Overviews appeared on a stated share of the saved panel; and enquiries from those pages changed by a stated amount, or did not change. We cannot isolate the cause.” The figures in this example are invented to show the shape of the report.

Common mistakes

• Dividing ordinary web clicks by AI impressions and calling it “AI Overview CTR”.
• Averaging page CTRs.
• Comparing months with different seasonality.
• Treating every lost click as a lost customer.
• Reporting traffic and hiding the outcome that did not fall.

Questions people ask

Did AI Overviews cut my traffic?

You can describe a pattern and its plausible causes. Proving the cause needs a comparison you rarely have.

Should I stop writing informational pages?

Not on the strength of CTR alone. Ask whether the page serves a task that still needs a visit, and whether being cited helps.

What should I track instead of CTR alone?

Qualified outcomes, citation and mention in the answers that matter, and the pages those answers cite.

Use the ROI workflow for cost and outcome modeling, and the content guide for improving the answer itself. Preserve the export and settings so the next review can repeat the comparison.