AI Mode: The Study That Contradicts Google on Clicks and UX
Summarize this article with AI
The Study That Puts a Number on the Feeling
I look at 15 sites a week. The same question keeps coming up. Should we invest in AI Mode? One number was missing. Here it is.
1,444 volunteers. 1,100 tracked for seven days. A browser extension recorded every search, every click, every screenshot of the results pages. The study is signed by researchers from the University of Pennsylvania and Northeastern University. It was published in August 2026. Search Engine Journal picked it up.
Google says people love AI Search and that traffic has never been better. The research paper says otherwise.
A field experiment, not a declarative survey. Randomized. Controlled.
The finding comes down to one sentence: Google’s AI features reduce clicks to sites and degrade user expérience. Not an opinion. A measurement.
What that changes for an e-commerce site.
Avant d’entrer dans les trois configurations, posons l’échelle réelle de l’étude. Un funnel limpide : 1 444 volontaires recrutés, 1 100 suivis pendant sept jours, 956 questionnaires finalisés.
Le parcours de l’étude
1 444 recrutés, 1 100 suivis sept jours, 956 questionnaires exploités.
Three Groups, Three Different Expériences
The researchers ran a randomized trial in real conditions. 1,444 volunteers recruited. 1,100 tracked by a browser extension that recorded every search, every click, every Google results page. At the end, 956 participants answered the questionnaire.
Three groups. Three configurations.
| Group | Configuration | Measured effect |
|---|---|---|
| Normal Google Search | Google with AI Overviews when Google shows them | Natural behavior |
| Forced without AI | AI Overviews hidden, AI Mode redirected to classic Google | Effect of AI Overviews |
| Forced AI Mode | All searches routed to AI Mode | Isolated AI Mode effect |
The experiment hit an obstacle. Google changed the HTML of its AI Overviews mid-test. Result: the no-AI group only had its AI Overviews hidden 90% of the time on day one. Then 0%.
In the end, 51.1% of AI Overviews were hidden. Median: 50%. That weakens the method. It doesn't kill it.
Imperfect. Measured.
Fewer Clicks, More Frustration
The numbers are clear. AI Mode reduces clicks to sites. It keeps the user inside Google search longer, and sends them to a website less often than a classic search.
The group forced into AI Mode reported less satisfaction with Google Search. Some went elsewhere. DuckDuckGo. Bing. Yahoo.
Why click, anyway? The answer is already in the box. The user got what they were looking for without leaving Google.
AI Mode keeps users in the Google search expérience longer and sends them to websites less often than a classic search.
In the no-AI group, clicks to sites increased. Removing AI Overviews kept performance steady. Measurable performance held.
Bottom line: without AI, satisfaction holds. With it, outgoing traffic falls.
Fewer clicks. Degraded UX. Departures to other engines.
Why Google Says the Opposite
Google reports record traffic. The study notes a drop in outgoing clicks. Who's right?
Both. Google counts interactions inside its engine, not visits to your site. Impressions go up. Outgoing clicks, they go down.
That's where the confusion comes from. A user happy with AI Mode stays on Google. Google sees a win. Your e-commerce site sees a blocked tunnel.
No secret: just two opposing goals. Google wants to keep the user. You want the user to come to you.
In 2026, SEO is no longer about raw visibility. It's about the outgoing click. An impression doesn't make a sale.
Your site can be cited in AI Mode. It can also appear in an AI Overview. Without a click, the citation doesn't pay your servers.
What I See on E-commerce Sites
For my e-commerce clients, the pattern repeats. Catalogs of 2,300 pages. Semantic clusters that capture informational queries. AI Mode aggregates the answers. The click stays absent.
Organic traffic stagnates. Average order value goes up, but qualified sessions become scarcer.
Architecture is the blocker. The content holds up.
A site that answers a question in a blog post no longer gets the visit. AI Mode synthesizes it. The reader stays on Google.
That's the negativity bias the DOSE framework talks about, taught by Guillaume Attias at BMO Academy. Documented discomfort draws attention. It gives credibility to the trade-off.
Liking AI Mode or not is secondary. The question that matters: does it send customers?
The Trade-off: AI Visibility or Real Traffic
Should you appear in AI Mode? Yes, if your business model lives on brand awareness. No, if you sell online. When in doubt: measure.
AI visibility and real traffic aren't the same column in your exports. An answer cited in AI Mode can bring zero clicks. A classic organic page, well positioned, can bring the sale.
What you gain by doing this: audit your pages that trigger AI Overviews. Look at outgoing clicks. Not impressions.
A page that gets impressions in AI Mode without any click works for Google, not for you.
Build clusters that answer transactional intent, not just informational. The click itself turns into a sale.
The method stays the same. I'll say it again: what matters is the click that sells.
What You Measure Is What You Optimize
Forget impressions. Look at outgoing clicks.
- Identify the queries where AI Mode cannibalizes your pages.
- Filter your exports by outgoing click, not by average position.
- Build clusters that answer purchase intent, not just the question.
- Compare real traffic before and after AI Mode integration in your target SERPs.
We stopped producing content for informational questions. We restructured. We invested the next $8,000 in the right place. +290%. Without ads.
Nothing magic. Just a trade-off.
The University of Pennsylvania and Northeastern study confirms what I see in the field. AI Mode moves the click rather than replacing it.
Does your SEO aim to rank or to sell?
Audit Your Clicks Before Investing in AI
During the audit, I measure what AI Mode actually cannibalizes on your pages. I show you outgoing clicks, not impressions.
Book a strategic call — 45 minFrequently Asked Questions
Will AI Mode replace SEO?
No. The trade-off shifts: outgoing clicks fall on informational queries. Transactional SEO keeps selling. Look at the right signals.
How do I measure the impact of AI Mode on my traffic?
I compare organic sessions page by page, filter by source, and look at outgoing clicks. Impressions earn nothing. A cited answer without a click earns nothing either.
Are the study's data reliable?
Yes. 1,100 randomly selected people, a browser extension, SERP screenshots. Google changed its HTML, AI Overview masking works less well. The method holds.
Should you block AI Mode via robots.txt or noindex?
Not necessarily. Measure cannibalization on your transactional pages. Then decide: visibility without clicks, or real traffic.
What does negativity bias mean in this context?
Documented discomfort draws attention. It gives weight to the trade-off. It's a tool from the DOSE framework, taught by Guillaume Attias at BMO Academy. I use it to decide.
