Are AI Overviews actually reducing clicks?

On informational queries, the pattern of steady or rising impressions alongside falling clicks is widely reported and easy to verify in your own data. The honest position is that you cannot attribute it from Search Console alone, because a ranking drop looks identical in the numbers. What you can do is check whether the affected queries are question-shaped and whether an AI Overview appears when you search them yourself. Commercial and local queries are affected far less than informational ones.

Can I see when ChatGPT or another assistant sends me traffic?

Partly. Direct referrals from some assistants show up in GA4 as referral traffic, so you can see the visits that came straight from a chat. What you cannot see is the much larger group who read a recommendation, closed the chat, and searched your name later. That arrives as brand organic search and is indistinguishable from a returning customer. Treat referral traffic from assistants as the visible tip of a larger effect.

Is a citation without a click worth anything?

It is worth something, but not something your search report can count. Being named in an answer puts your business in front of someone at the moment they are choosing, which is the same job the click used to do. The evidence usually turns up later and elsewhere: a brand search, a direct visit, or an enquiry from someone who already knows what you do.

Should I block AI crawlers to protect my traffic?

That trade is worth thinking through rather than reacting to. Blocking removes you from the answers that are increasingly where people choose a supplier, and it does not bring the click back, because the answer gets written from someone else's page instead. The more useful question for most small businesses is whether the pages being summarised are the ones you would want quoted.

How do I tell an AI Overview from a ranking drop?

Look at impressions. A ranking drop usually takes impressions down with the clicks, because you are being shown less often. An answer absorbing the click leaves impressions steady or rising while the clicks fall away. Then check the queries themselves: if they are questions rather than commercial or local searches, an AI Overview is the likelier explanation.

Does any of this change what I should write?

Not really, which is the reassuring part. The pages that earn a citation are the same pages that used to earn a click: specific, clearly written, and genuinely about the problem rather than about you. What changes is how you measure them, since the payoff now shows up as a brand search rather than a non-brand one.

How AI is Changing Organic Search

Hero: an AI generated answer sitting above the first blue link, with the link pushed down the page

Your impressions are up. Your clicks are down. Nothing about your website changed.

That combination used to mean one thing: you slipped down the page. It now has a second explanation, and telling the two apart matters, because they call for completely different responses.

This guide assumes you have already split your search traffic into brand and non-brand. If you have not, start with Brand vs Non-Brand in Organic Search, because everything below is about what is happening to the non-brand half.

What AI Overviews changed

The answer now appears on the results page, and it takes the non-brand click with it.

An AI Overview is Google answering the question itself, at the top of the page, before any link. It appears most often on exactly the queries that make up your non-brand half: the informational ones. "How much does a retaining wall cost", "how often should I service a heat pump", "what is the difference between X and Y".

When that happens, your page can still be cited in the answer. You still get the impression. The searcher reads the answer and never clicks.

Brand queries are largely untouched by this, because someone searching your company name is not asking a question Google can answer for them. They want you, and they click through to you. So the two halves move apart.

Figure 1. The scissors

Non-brand impressions hold up or grow while non-brand clicks fall away. Brand clicks carry on as if nothing happened.

Non-brand impressions Non-brand clicks Brand clicks

Illustrative figures. Read carelessly, the last month looks like a brand getting stronger: brand share is up. Nothing about the brand changed. The non-brand half was absorbed by the answer at the top of the page.

This is the pattern worth learning to recognise, because it is genuinely confusing. Two things that normally mean good news, a rising brand share and a growing impression count, can both appear while your actual reach into new customers is shrinking. The tell is the gap between non-brand impressions and non-brand clicks. If it is widening, you are being shown and not chosen.

The same widening gap has a second, more ordinary cause: you rank on page two, or your title and description do not match what was asked. Both are worth fixing and both look identical in the totals. That is why the split is the first cut, before any of the rest of it.

The LLM blind spot

Discovery is moving to places that never appear in your Search Console at all.

The question that never reaches your report, and the three names that come back

Off-platform discovery

The question you never see

The bigger shift is happening outside Google. When someone asks ChatGPT, Claude or Gemini "who should I use for commercial landscaping in Auckland", that is a non-brand query in every sense that matters to your business. It never touches your Search Console. There is no impression, no click, no row in any report you own.

What often happens next is the interesting part. They get two or three names, and they Google one of them. That search arrives in your report as brand.

Figure 2. Where the discovery went

The non-brand step moved off-platform. It arrives in your report wearing the brand label.

Step 1

A need, no name

"Who does commercial landscaping in Auckland?"

Step 2

Asked an assistant, or read the AI Overview

Two or three businesses are named.

Step 3

Googles the name

"Lou's Landscaping reviews"

Counted as brand

Step 4

Clicks, and enquires

Your report credits brand search with the win.

Counted as brand

The discovery that actually won this customer happened at step 2, where you have no measurement at all. Your report sees a brand search and a conversion, and reasonably concludes your brand is strong.

This changes what brand search means. It used to be a clean read on existing reputation: people who already knew you, coming back. Increasingly it is also the arrival lane for people who learned your name thirty seconds earlier from a machine.

You cannot separate those two groups inside Search Console, and you should be suspicious of anyone selling you a tool that claims to. What you can do is treat brand search volume as a recommendation signal rather than a performance one. If brand searches climb while your advertising, your PR and your customer count are all flat, something is putting your name in front of people. That is worth investigating rather than congratulating yourself for.

A practical consequence: the content that gets you recommended by an assistant is the same content that used to earn non-brand clicks. Clear, specific, genuinely useful pages about the problem you solve. The click may not arrive any more, but the citation does, and the brand search follows it.

What to do about it

Four changes to how you read the report.

1 Watch the non-brand impression to click gap.
Widening means you are being shown and passed over. That is the single most useful number to add to a monthly search report.
2 Check whether the affected queries are question-shaped.
Informational queries are where AI Overviews appear. Commercial and local queries are affected far less. Search a few of them yourself and look.
3 Treat brand volume as a recommendation signal.
Rising brand search with flat advertising and a flat customer count usually means something is putting your name in front of people. Go and find out what.
4 Keep writing the specific, useful pages.
The content that earns a citation is the content that used to earn the click. The click may not arrive now, but the citation does, and the brand search follows it.

The takeaway

Being shown and being chosen have come apart, and your report only counts one of them.

AI Overviews and assistants have made the brand and non-brand split more useful and harder to read at the same time. Non-brand clicks are getting scarcer without non-brand demand disappearing, and some of that demand now arrives wearing the brand label.

The businesses that will read this correctly are the ones already splitting the two, watching the gap between being shown and being chosen, and treating a rising brand share as a question rather than an achievement.

In Clearly: the Search Console page in every report splits brand from non-brand automatically and shows impressions next to clicks for each half, so a widening gap is visible without you going looking for it.

How AI is Changing Organic Search

What AI Overviews changed

The LLM blind spot

The question you never see

What to do about it

The takeaway

Brand vs Non-Brand in Organic Search

Clicks, Impressions, CTR & Position: A Worked Example

Why Search Console and GA4 Don't Match

Content Grouping in GA4