AI Overviews Are Cutting Clicks: Search Console Won't Name the Keywords

The position is still 3, and clicks alone have fallen to 60% of last month's. You can get as far as guessing the cause is AI Overviews. The trouble starts after that: Google Search Console doesn't say which keyword lost the clicks. Searching every one of them is out of the question, so you narrow to a handful using two things — the size of the drop and revenue.

Contents

Rankings hold steady while only the clicks fall
Google Search Console can't identify which keywords show an AI Overview
Decide which keywords to defend by drop size and revenue
How RevenueScope helps
FAQ

Summary / References / Related articles

TL;DR#

In data where Seer Interactive tracked 53 brands and about 5.47 million search keywords, the click-through rate on searches showing an AI Overview was 2.36%, against 3.82% on searches without one[7]
In Japanese too, a survey taken in April 2026 found an AI Overview on 46.9% of purchase-intent keywords[5]
The generative AI performance report answers only as far as impressions, so the way clicks fell can't be followed keyword by keyword[2]
Impressions and clicks inside an AI Overview are added into ordinary search performance and never counted separately[1]
Checking every keyword by eye is impossible, so cut the list down to the handful where falling clicks and large revenue overlap

1. Rankings Hold Steady While Only the Clicks Fall#

This change doesn't surface in a rank table. The position stays at 3 and clicks alone fall to 60% of the previous month. If the ranking had dropped you would have somewhere to start, but with the ranking unmoved no trace is left behind. There is no move to raise the ranking either, since it is already 3.

The background numbers are in. Seer Interactive, a US marketing firm, tracked 53 brands and about 5.47 million search keywords and found that the organic click-through rate was 2.36% on searches where an AI Overview appeared and 3.82% on searches where it did not (as of February 2026[7]). In the same data, the rate with an AI Overview has come back from 1.31% in December 2025, and the gap is narrowing. It has not closed. When Ahrefs measured informational keywords on Google in the US, the click-through rate at position 1 was about 34.5% lower where an AI Overview was present[6].

On the Japanese search side, there is a survey of 2,609 keywords taken in April 2026. An AI Overview — written as "AIによる概要" in Japanese — appeared on 46.9% of purchase-intent keywords[5]. It is a snapshot of a single point in time, so it can't say whether that is up or down, but it does show that they turn up on nearly half of even the keywords closest to purchase. No first-party data measuring how far the click-through rate fell in Japanese is available, so nothing more can be said than that a similar tendency is presumed in Japan.

A bar chart of how often an AI Overview appears, split by search intent. Searches to find out about something read 70.9%, searches to get something done read 66.0%, and searches to buy read 46.9%. Even on keywords closest to purchase, an AI Overview shows on nearly half. A single point-in-time survey of Japanese searches, April 2026

When the ranking holds and clicks fall, the first job is to separate two cases: the AI Overview took the answer and the click disappeared, or you were picked as one of the sources it cites. Same rank but fewer clicks: how to tell which covers how to tell them apart.

This article picks up after that separation is done. For the keywords that look affected, it covers only how to decide the order in which you defend them.

2. Google Search Console Can't Identify Which Keywords Show an AI Overview#

Your own Google Search Console figures alone cannot confirm, keyword by keyword, which keywords got an AI Overview and lost clicks because of it.

The first place you think of is the generative AI performance report. It covers exposure in AI Mode and AI Overviews, but as of the July 2026 help documentation the only metric it carries is impressions[2]. Neither clicks nor position appears in that report's figures.

Next you look at the ordinary search performance report. When a link to your site appears inside an AI Overview, those impressions and clicks are counted within "Web" type search performance[1]. No row is provided that pulls out the AI Overview portion on its own. What you are looking at was combined from the start.

Position works the same way. Even when an AI Overview contains links to several sites, it is counted as a single position within the search results[3]. The position shown is the value of the place the AI Overview occupies, and where inside it you were listed is not reflected.

A table of what Google Search Console can and cannot show about the impact of AI. Impressions inside AI answers can be seen. Clicks arriving from AI answers, the list of keywords that show an AI Overview, and the position held inside an AI Overview cannot be seen

Lay these three together and the range your own figures can settle is fixed. Impressions and clicks are totals that include the AI Overview portion, and position is where the AI Overview sat. Nowhere is there a column that carves out, keyword by keyword, how many of the lost clicks were down to the AI Overview. "AI Overviews cost us clicks" cannot be verified directly from your own raw figures.

On the sample data, which keywords that lost clicks still matter by revenue?

3. Decide Which Keywords to Defend by Drop Size and Revenue#

Since the figures can't identify it, what remains is to search and check with your own eyes. Search the keyword on Google and see whether an AI Overview appears. That part costs nothing.

What this confirms, though, is only the state at that moment. AI Overviews are summaries produced by generative AI[4], and they are shown only when the system judges them useful[1]. The same keyword brings one up at some times and not at others. What people see also shifts with their situation and the device they use, and the picture changes again once the month turns. Seeing nothing once is not proof that the keyword is unaffected.

There is also a problem to settle before that: you have to decide which keywords to look at. Searching several hundred keywords every month is unreasonable, and once through is not enough. The work of narrowing the list comes first.

There are two conditions for narrowing. Clicks have fallen against the previous period, and the keyword produces revenue. Only keywords that meet both go on the list you check by eye. Keywords that have fallen but have nothing to do with revenue can wait, and keywords with large revenue whose clicks have not fallen can be left alone for now. Picking which rankings to push works the same way: choose by the effect on revenue, not by the ranking itself (how to pick ranks worth chasing).

A scatter plot placing seven search keywords from the sample store on click drop along the horizontal axis and estimated revenue along the vertical axis. The two keywords in the upper right, where the drop is large and estimated revenue is large as well, are the first ones to check by eye

For the keywords you decide to defend, there are two directions to move in. If the AI Overview is taking the answer, aim to become one of the sources it cites (who AI Overviews actually cite). If you are already being picked as a citation, you are better off not breaking the shape of the article even though clicks have fallen (zero search clicks but chosen by ChatGPT). Which way to go is left to those articles. What this article covers is narrowing the list you make that call on down to a handful of keywords.

RevenueScope solution

What you need on hand to decide which keywords to defend is two columns: the period-over-period click change for each keyword, and the revenue that keyword produced. Google Search Console gives clicks by keyword, but revenue does not come out by keyword. That is where narrowing the list to check by eye stalls (why revenue by search keyword doesn't show up even with GSC connected).

RevenueScope fills that empty second column with estimated revenue. Estimated revenue and clicks are both shown with the rate of change against the previous period, so sorting the keywords that have fallen by estimated revenue, largest first, takes the handful to check by eye off the top. Estimated revenue is an approximation worked back from search-driven revenue, not settled revenue. It covers Google search data, and reflection lags by two to three days.

What the shortlist looks like#

Take a fictional cosmetics store and compare four keywords.

Keyword	Clicks vs. prior period	Estimated revenue	Revenue vs. prior period
moisturizer for sensitive skin	-42%	¥380,000	-31%
how to use a first essence	-55%	¥240,000	-40%
skincare order	-48%	¥20,000	-45%
affordable toner	-6%	¥310,000	+2%

Note: these exact figures aren't what comes up on screen. The table above is a worked example rounded for explanation, while the demo screen runs on the sample store's data.

Reading it is simple. The top two keywords have lost between two-fifths and more than half of their clicks, and their estimated revenue is large as well. Checking those two by eye is enough. The third has fallen just as hard but its revenue is small, and the fourth holds its large revenue with clicks barely down. Out of several hundred keywords, the ones you actually go and search come down to a handful.

The job of updating those two columns every month carries on for as long as AI Overviews keep spreading. RevenueScope does that updating automatically.

In the sample store, which keywords still earn revenue after clicks dropped?

FAQ#

Frequently asked questions#

Q. Will RevenueScope tell me which keywords have an AI Overview?

A. Whether one appears is something you confirm by looking at Google's search results yourself. What RevenueScope provides is the material for deciding which keyword to start that checking from: the period-over-period click change and the estimated revenue for each search keyword.

Q. How far off from actual revenue is estimated revenue?

A. Estimated revenue is produced by applying revenue per search-driven session to that keyword's click count. It is built to come out on the low side, and it is not settled revenue. Rather than putting the absolute amount into a report, use it to compare keywords against each other.

Q. How often should the by-eye check be done?

A. Once a month is the guide for the keywords you have decided to defend. The same keyword brings up an AI Overview at some times and not at others[1], so don't reuse a single result for long. Keep the list down to a handful of keywords and a monthly check doesn't become a burden.

Summary#

In search results where AI Overviews have spread, you can't go and defend every keyword. In data where Seer Interactive tracked 53 brands, the organic click-through rate was 2.36% on searches showing an AI Overview and 3.82% on searches without one[7]. In Japanese too, a survey found them on 46.9% of purchase-intent keywords[5].

And which keywords your clicks fell on cannot be confirmed keyword by keyword from Google Search Console figures alone[1][2][3]. The only way to check is to look at the actual search results, and that can't be done across the board.

So the list gets narrowed first. Whether to defend is decided not by ranking but by two things: whether clicks are down against the previous period, and whether estimated revenue is large.