38% fewer clicks, satisfaction unchanged: the AI Overviews paradox
Summarize this article with AI
First randomized study: 38% lost clicks, identical satisfaction
A client calls me mid-March. B2B e-commerce, 47,000 organic sessions per month in January. 32,000 in February. He changed nothing on his site.
The problem?
Google’s AI Overviews.
And this isn’t a feeling. It’s measured.
A study published this month on SSRN by the Indian School of Business and Carnegie Mellon University proves it with unprecedented rigor: AI Overviews reduce organic clicks by 38% on queries where they appear, according to the experiment led by Saharsh Agarwal and Ananya Sen. First randomized study in real conditions. Not scraping. Not dubious correlation. Real controlled experimentation.
1,065 Chrome desktop users in the United States, recruited via Prolific. Three groups:
- Control group: normal Google
- « Hide AIO » group: Chrome extension that masks AI Overviews in real time
- « AI Mode » group: systematic redirect to Google’s AI mode
Duration: two weeks per participant, between January and February 2026. Protocol pre-registered with the AEA RCT Registry before data collection. No confirmation bias possible.
Raw result:
When AI Overviews appear (42% of queries in the study), the organic click rate drops from 0.61 clicks per search to 0.38. That’s a 38% decline.
Zero-click searches explode: from 54% without AI Overview to 72% with.
And now, the twist.
Stated user satisfaction? Identical. Perception of result quality? Identical. Ease of finding information? Identical.
Removing AI Overviews changed nothing about the subjective expérience users reported.
This isn’t a bug. It’s a complete business model redesign.
Dopamine arrives before the click: what the DOSE framework teaches us
I’ve worked with the DOSE framework since 2022, taught by Guillaume Attias in the BMO Academy training. Four neurological levers that govern attention and action online:
- Dopamine: anticipation of reward
- Oxytocin: social connection, trust
- Serotonin: status, recognition
- Endorphin: immediate pleasure
Until now, we thought dopamine was released after the click, when the user found their answer on your page.
Agarwal and Sen’s study proves otherwise.
The cognitive reward is delivered in the SERP. Before the click. The AI Overview gives the answer. The user gets their dopamine hit. The loop closes.
No need to visit your site.
That’s why satisfaction stays the same: the brain got what it was looking for. The question « What temperature to roast a chicken? » finds its answer at 350°F for 1 hour 20 minutes. No need to click on a cooking site.
Among my clients, I’ve observed this exact phenomenon since November 2025. Short informational queries lose 30 to 45% of their traffic once an AI Overview triggers. Bounce rates on those pages don’t move—because users still clicking are those seeking more than the synthetic answer.
In other words: traffic drops, but its quality rises.
That’s not a consolation. That’s a metric shift.
72% of zero-click searches: the click is no longer the norm
The numbers are clear.
Without AI Overview: 54% of zero-click searches.
With AI Overview: 72%.
An absolute majority of queries no longer generates an outbound click.
I review 15 sites per week in audits. All show me the same pattern since January 2026: traffic loss on informational queries with weak intent, stability or gains on transactional or complex queries.
Recent client example (March 2026):
B2B SaaS site, 120,000 organic sessions per month in October 2025. 87,000 in February 2026. CMO panic mode.
We dig in.
340 blog pages lose an average of 38% of traffic. All « How to do X » guides with short answers (300–600 words). Google displays an AI Overview on 68% of these queries.
But.
47 product pages gain +12% over the same period. 15 detailed comparison pages gain +19%. Demo landing pages gain +23%.
Valuable traffic increases.
Pure informational traffic disappears.
It’s not a crisis. It’s a redistribution.
According to Agarwal and Sen’s study, the effect is strongest when the AI Overview appears at the top of the page (85% of cases in their sample). In this configuration, removing the AI Overview nearly doubles outbound clicks. When the AI Overview is lower, the impact becomes negligible.
Position = everything.
And you don’t control the AI Overview position. Google does.
Paid clicks stay flat: Google protects its revenue model
Here’s what the study doesn’t trumpet in its title, but what’s in the data.
Clicks on sponsored ads remain stable whether or not an AI Overview is present.
Zero measurable impact.
That’s no accident.
Google has architected its AI Overviews to cannibalize organic traffic, not paid traffic. The cognitive reward arrives via the AI answer, but if the user wants to buy, they click on an ad or a transactional organic link.
I’ve observed this with my clients since December 2025. Google Ads campaigns on « buy X », « best X 2026 », « X cheap » queries lose nothing. Sometimes they gain, because parasitic organic clicks (users in pure research mode) disappear.
Cost per conversion drops.
Conversion volume stays flat or rises.
Google doesn’t kill its own revenue engine. It kills non-transactional organic traffic. The kind that never made Google any money anyway.
It’s brutal. It’s rational.
And it upends 15 years of SEO strategy built on session volume.
What SEO strategy in 2026 faced with AI Overviews?
I’m not selling you a method. I’m showing you what works for my clients in the last four months.
1. Stop producing short informational content
If your page answers a simple question in 400 words, the AI Overview will cannibalize it. No point fighting.
Either you delete those pages and redirect their authority to transactional content.
Or you expand them: 2,000+ words, multiple use cases, proprietary data, comparison tables. The AI Overview can’t summarize everything.
E-commerce client (February 2026): we removed 120 cannibalised « quick guide » pages. Redirected to 18 comprehensive guides (2,500–3,800 words each). Result after 6 weeks: +17% overall traffic, +31% on strengthened pages.
2. Bet on transactional and commercial intent
« Buy », « comparison », « review », « vs », « best » queries hold their ground. The AI Overview appears less (28% vs 42% across all query types per my observations). And when it does appear, it often links to external sources.
Your product pages, detailed comparisons, conversion pages: that’s where you invest in 2026.
3. Structure to be cited in the AI Overview
If you can’t avoid the AI Overview, become its source.
Tables, numbered lists, unique numerical data, structured markup with Schema.org. The AI Overview cites its sources as clickable links in 34% of cases (observation across 240 test queries in March 2026).
SaaS client (January 2026): we restructured 22 articles with comparison tables and FAQ Schema. 9 are now cited in AI Overviews. These 9 pages lose 29% direct traffic, but gain authority: +12% natural backlinks in 8 weeks.
4. Measure conversion, not sessions
That’s the real revolution.
If you still measure SEO in « organic sessions », you’ll panic. If you measure in « qualified SEO leads » or « SEO-attributed revenue », you’ll see it holds.
Traffic drops. Conversion rate climbs.
Among 11 of my clients, average organic conversion rate moved from 1.8% (October 2025) to 2.4% (March 2026). Organic conversion volume dropped 6% over the same period. Not zero. Not catastrophic.
We adapt. Like always.
The study’s limitations: desktop, Chrome, United States
Agarwal and Sen’s study is solid. But it has acknowledged limits.
1. Desktop only
The experiment ran on Chrome desktop. Yet 63% of Google traffic comes from mobile (Statcounter data, March 2026). AI Overviews take up even more screen real estate on mobile. The effect is probably higher than 38% on mobile.
2. Active Chrome users
Participants needed a minimum browsing history. These aren’t « average » users. They’re more accustomed to Google interfaces, potentially more inclined to trust AI Overviews.
3. January–February 2026
The study is two months old. AI Overviews evolve fast. Google tweaks triggers, formats, cited sources. These figures are a snapshot, not gospel truth.
4. No measurement of commercial intent
The study doesn’t segment by query type (informational vs transactional). We know the impact varies. My own observations show a 15 to 20 point gap between the two.
All that said: 38% is an average. Your site might lose 50%. Or 15%. It depends on your query mix.
Traffic and user value are now disconnected
For 20 years, we believed more traffic = more value.
That held when Google returned 10 blue links.
It’s no longer true when Google gives you the answer before you click.
Users are satisfied. They get their dopamine. They don’t click.
Your site loses traffic. Your SEO revenue can hold steady.
It’s not a crisis. It’s a business model transformation.
Those still measuring in sessions will panic.
Those measuring in conversions will adjust.
I’ve built SEO systems for 650+ clients since 2016. I’ve seen Google Panda, Penguin, featured snippets, People Also Ask, Core Updates. Each time, we thought it was the end.
Each time, we adapted.
AI Overviews don’t kill SEO. They kill low-value informational SEO. The rest resists. The rest strengthens.
The real question isn’t « How do I recover those 38%? »
It’s « How do I extract more value from the 62% that remains? »
Are your transactional pages converting at 2% or 4%? Are your comparisons cited in AI Overviews? Are your semantic clusters pushing long-tail queries the AI doesn’t cover?
Do you have the answers for your site?
Audit: does your site withstand AI Overviews?
I measure the real impact on your strategic queries. I show you which pages hold, which sink, and how to restructure your architecture. First audit free, on video call, using your own Search Console data.
Book a strategic call — 45 minFrequently Asked Questions
Do AI Overviews really reduce organic traffic by 38%?
Yes, according to Agarwal and Sen’s randomized study (Indian School of Business + Carnegie Mellon, 2026) on 1,065 Chrome desktop users. Organic clicks drop from 0.61 to 0.38 per search on queries where an AI Overview appears.
Why does user satisfaction stay the same despite fewer clicks?
Because the cognitive reward (dopamine) is delivered in the SERP, before the click. The user gets their answer via the AI Overview. Their brain considers the task complete, even without visiting an external site.
Do AI Overviews also impact paid ad clicks?
No. The study shows paid clicks remain stable. Google designed AI Overviews to cannibalize organic informational traffic, not its ad model. Transactional queries show more resistance.
How do I adapt my SEO strategy to AI Overviews in 2026?
Stop making short informational content. Focus on transactional and commercial intent. Structure your content to be cited in AI Overviews. Measure conversions, not sessions. Traffic drops, quality rises.
Is the AI Overviews study valid for mobile?
The study covered Chrome desktop only. On mobile (63% of Google traffic), AI Overviews consume even more screen space. The impact is likely higher than 38%, but no randomized study has measured it yet.
