AI Overviews: Google Tests Clicks, but Revenue Tells a Different Story
Summarize this article with AI
A client calls me. He lost 40% of traffic. The reason isn’t what you think.
6,800 organic sessions per month. Catalog of 14,000 products. Clean growth for 3 years. Then, in 6 weeks, -40%. No manual penalty. No technical bug. Just a Google update on the AI Overviews side.
The site owner reads Google’s official statements. Liz Reid, Search lead at Google, talks about « bounce clicks. » Visits that click, read a fact, leave in 8 seconds. Google deems these clicks low-value. It removes them with the AI overview. The argument is attractive. But the client’s numbers don’t confirm it.
I look at the pages hit hardest. Comparison guides. Technical specs. Concrete questions. Nothing that looks like bounce traffic. Google just decided the answer is better without the site. And the site takes the hit.
38% fewer organic clicks. A randomized experiment that unsettles.
A research paper led by the Indian School of Business and Carnegie Mellon University settled it. 1,065 American participants. Three groups split by a Chrome extension. Group 1: Google Search normal with AI Overviews. Group 2: without AI Overviews. Group 3: full AI Mode. Net result.
38% fewer organic clicks when the overview is present. 33% more zero-click searches. And most striking: removing the overview degrades neither user satisfaction, nor perceived quality, nor ease of finding information. In other words, the AI doesn’t improve the expérience. It just captures attention.
« The problem isn’t the content. It’s the display surface. Google monopolizes it. »
That same week, Liz Reid reaffirms on Bloomberg that the missing « bounce clicks » did no harm to publishers. No public data to prove it. A narrative. Not proof.
Alphabet +16%, Microsoft +14%: Quarterly earnings don’t mention bounce clicks.
While Google reassures publishers about click value, Alphabet posts results. Advertising revenue up 16% year-over-year. $76.47 billion for the quarter. On Microsoft’s side, Search & News Advertising jumps 14%. Two giants, two record quarters.
Surprise. Because if AI Overviews reduce organic clicks, the ad machinery doesn’t weaken. Better: it visibly flourishes. Why? Because an AI overview isn’t ad-free space. It embeds Shopping ads, sponsored links, formatted answers. The user stays in the ecosystem. Fewer external destinations, more ad touchpoints.
The two narratives advance in parallel. One talks about editorial quality, bounces, satisfied users. The other talks revenue, ad margin, SERP stickiness. This isn’t a bug. It’s economic architecture.
What the DOSE framework taught me about gaps between promise and reality.
The DOSE framework, taught by Guillaume Attias at BMO Academy, breaks down four phases: Data, Observation, Strategy, Execution. But it emphasizes one underrated step: Expected vs Surprising. It’s not the expected that creates advantage. It’s the counter-intuitive.
Here, the expected: AI Overviews would shrink organic traffic. The surprising: despite that drop, search engine revenue explodes. The gap is there. And in that gap, your next edge lives.
When I model a semantic cocoon for a client, I no longer set click volume as the only KPI. I look at brand position, entity coverage, ability to answer conversational questions without depending on the blue link. A well-structured page might stop receiving the immediate click, but it feeds the entire typed mesh. And the site wins at 6 months, when Google adjusts its weights.
Build for the question, not the link.
I observe a clear shift at my clients. Pages that survive AI Overviews best are structured as question-answer, with FAQ, HowTo, QAPage schema. No padded text. No sales tunnel. A <question> tag and a factual, sourced answer in under 60 words.
A B2B site with 420 articles pivoted in 4 months. We restructured 80% of the catalog into question-answer blocks. Result: featured snippets dropped 30% (expected—Google doesn’t need to show them). But long-tail traffic climbed 47%. Because complex questions, ones requiring nuanced answers, aren’t digestible by an AI overview. They need a click. And the site was there, structured, fast, marked up.
That’s the mechanism. You don’t fight the AI Overview. You bypass it through genuine need.
The quarterly lesson: when Google monetizes better, so can you.
Alphabet doesn’t just monetize clicks. It monetizes attention on the SERP. Every second spent on an AI overview is a second the user doesn’t leave Google. And Google charges advertisers for that attention.
For a merchant site, it’s a constraint. But it’s also an opening. If the user seeks a comparison, a brief answer in the overview doesn’t close their need. It starts it. When the decision gets complex, the click returns. But only if the page is ready: entity-structured, linked in a coherent internal mesh, indexed within 24 hours.
My e-commerce client went all-in there. We cut 1,200 empty pages. We concentrated effort on 380 core pages, meshed in a cocoon. Audience bounced back in 5 months. +29% conversions. Organic traffic never hit pre-AI Overviews levels. But revenue did—and beat it.
Is your site structured for the user who doesn’t click… before they click?
I build systems that capture intent, not pages that await the click. That’s the 2026 flip. AI Overviews aren’t a blip. They’re redefining the information architecture of the web.
When I model a semantic cocoon today, I don’t reason in keyword volume anymore. I reason in entity mastery. In coverage of non-obvious questions. In stacking trust on a SERP that shrinks room for blue links.
And it changes everything. It changes the mesh. It changes production. It changes ROI. The click must be earned. Not because Google is generous. Because your page is the only one that answers what the overview couldn’t say.
And you? Will your next audit start with clicks or with questions?
A live audit to see where Google harvests your questions without clicking
I don’t sell you the method. I show you the pages. First call: 45 minutes of live audit on your site. We identify which questions the AI Overview can’t yet digest, and the path to turn them into traffic.
Book a strategic call — 45 minFrequently Asked Questions
What does Google mean by a « bounce click »?
Google defines a bounce click as a fast click on a page to grab a fact, then leave immediately. According to Liz Reid, AI Overviews remove these low-value visits. But Google has released no data to back this distinction.
Why do ad revenues grow despite falling organic clicks?
AI Overviews keep the user longer on the SERP. Google can embed Shopping ads, sponsored links, and AI-generated answers there. Fewer outbound visits, but more captive ad impressions, driving revenue growth.
How can a site adapt to AI Overviews?
Structure content into factual question-answer blocks. Use schema markup (FAQ, HowTo). Target questions too complex for the AI to summarize in a paragraph. Long-tail traffic and conversions hold better than high-volume pages.
Is the Indian School of Business study reliable?
Yes. It’s the first real-world randomized experiment with 1,065 participants. Not correlative—causal. Results (-38% clicks, unchanged satisfaction) are robust, though the sample is US-only.
Can a semantic cocoon protect my traffic from AI Overviews?
Not protect, but reposition. It shifts your site toward higher-value intents, less covered by overviews. Good semantic mesh improves discovery and dwell time—both signals Google leverages, even in the AI era.
