“Traffic’s Down Since The AI Answers Showed Up, And There’s Nothing We Can Do”
Somebody says a version of this in every monthly review now. Rankings held. Rankings even improved. Clicks went down anyway.
I’ve sat in that meeting. The mood that follows is the worst part — a quiet agreement that search is weather now, and you don’t optimize weather.
That was my read too. When a typhoon comes through, the quality of your umbrella stops mattering much.
But somebody actually pulled the numbers on 53 brands and 5.47 million queries, and it turns out we didn’t all get soaked the same amount.
Inside the same search result, the clicks split about 2.2 to 1. So let’s look at what decided which side you land on.
How Many More Clicks Does A Cited Brand Get From An AI Overview?
On the same search result, brands cited inside the AI summary got roughly 2.2x the clicks of brands that weren’t cited — a +120% gap per impression.
The numbers come from Seer Interactive, a US digital marketing agency, published April 24, 2026.
is the AI-written summary Google drops at the top of the results page (AIO for short). Which, for you, means part of the page now answers the question before anyone reaches a link.
The sample is the agency’s own client roster: 53 brands, 5.47 million queries, 2.43 billion organic impressions and 296.9 million paid ones. The window runs January 2025 through February 2026 on measured data, with March 2026 estimated.
The design is the useful bit. They lined up three states of the same query:
- no AI summary at all
- an AI summary that cites your brand
- an AI summary that doesn’t
And here’s what came out, counted as clicks per million organic impressions:
- no AI summary: about 33,500 clicks
- AI summary, cited: about 20,743 clicks
- AI summary, not cited: about 9,445 clicks
Cited is roughly 2.2x uncited. Same query, same page, same moment.
The Thing That Split Them Wasn’t Rank
Let’s put the ugly number on the table too. Even when you’re cited, you’re down 38% against queries with no AI summary. The total pool shrank, and no amount of framing fixes that.
But whether Google shows an AI summary on your query isn’t yours to decide. Google decides it.
What’s yours is whether your name is inside the thing once it shows up.
And that — the part you actually control — is where the 2.2x sits. That’s the whole finding.
“AI Overviews cut our clicks” has been a story about the total until now (ranking first and getting no clicks anyway). What this data adds is the distribution underneath it.
The pool got smaller, and then cited brands took a bigger slice of what was left while uncited brands took a smaller one. That’s a different problem with a different fix.
It also points the same direction as the finding that citation beats position — the reason there’s room below rank 10 that a rank obsession can’t see.
Citation Showed Up In The Paid Ads Underneath, Too
The same study reports paid click-through rate, averaged across all of 2025:
- no AI summary: 19.75%
- AI summary, cited: 15.74%
- AI summary, not cited: 11.19%
So whether you were cited in the organic summary tracked with how the ads below it performed. Your ad, your budget, your bid — and a variable sitting outside all three.
This is correlation, obviously. Nobody ran an experiment; these were observed together.
Still, there’s a plain reading available: people who just saw your name in the summary are readier to click your ad underneath. Recognition, then the click.
I hear “our paid efficiency is slipping” a lot lately. Maybe some of that isn’t happening inside the ad account at all.
Three Caveats You Have To Carry With The Number
If you’re taking this into a meeting, take the error bars with it.
- Seer Interactive sells search and AI-search services, so “invest in AI visibility” is a conclusion the publisher profits from
- The sample is 53 of its own clients, which leaves real bias by industry and company size
- It’s US-centric data, not a measurement of your market unless that’s your market
One more thing worth watching: the timeline wobbles.
Organic click-through rate for cited brands fell from 3.04% in Q1 2025 to 0.95% by Q4. Then it came back to 2.36% in February 2026.
So the behavior inside AI summaries hasn’t settled. I wouldn’t harden a single year of this into doctrine.
Add A “Cited” Column Next To Your Rank Table
Clicks from search dropped overall once AI summaries arrived. That part you absorb.
What you can move is the split afterward. On the same search result, cited brands and uncited brands sat 2.2x apart.
Watching rank will never show you that gap, because a rank table has no column for whether the model said your name.
So I’d narrow it to two moves.
First, check whether an AI summary actually appears on your main queries. If it doesn’t, this whole discussion is somebody else’s problem for now.
Then, on the queries where it does appear, count what share of them name you — on a schedule. One observation tells you nothing about direction, so ask the same questions monthly, under the same conditions.
Maybe I’m wrong about how fast this lands outside the US. But once that column exists, “rank went up and traffic went down” finally has somewhere to be explained.
Pick five queries that matter, run them, and count. That’s an afternoon.
Sources
- Seer Interactive, “AIO Impact on Google CTR: 2026 Update”, April 24, 2026, seerinteractive.com (53 of the agency’s own client brands, 5.47 million queries, 2.43 billion organic impressions and 296.9 million paid impressions; measured January 2025 through February 2026, with March 2026 estimated. On the same queries it compares three states — no AIO, AIO with citation, AIO without citation. Clicks per million organic impressions: about 33,500 with no AIO, about 20,743 when cited, about 9,445 when not cited, putting cited at +120% against uncited and -38% against no AIO. Paid click-through rate averaged across 2025: 19.75% with no AIO, 15.74% cited, 11.19% uncited. Organic CTR for cited brands: 3.04% in Q1 2025, 0.95% in Q4 2025, 2.36% in February 2026. Caveats: the publisher is an agency selling search and AI-search marketing services and has a commercial interest in this conclusion; the sample is limited to its own 53 clients, with possible bias by industry and company size; the data is US-centric and does not measure other markets.)
