THESYNTHETIC AGENT
The Synthetic Agent / Essay

Search Is Bigger.
Clicks Can Still
Get Smaller.

Google says AI is driving record search usage. A randomized field experiment found forced AI Mode sharply reduced external click-through. Both can be true.

Search is not dying. The click is losing its monopoly.
Here’s the point

Search can grow while website clicks shrink. There is no contradiction. AI lets people do more of the research, comparison and follow-up work inside the search interface, so the number of searches can rise while a smaller share of those searches need to become a visit.

Here’s the part I think a lot of people are missing.

Search is not dying.

The click is losing its monopoly.

Those are two very different things, and if your entire SEO report is still traffic, rankings and a green arrow, you are measuring the old interface.

Google can be right. Publishers can be right too.

Google says AI is expanding Search. In May, the company said AI Mode had passed one billion monthly users, AI Mode queries were more than doubling every quarter, and overall Search queries had reached an all-time high. Google later said AI Overviews had more than 2.5 billion monthly active users and that AI features were sending billions of clicks to websites every week.

I believe all of that can be true.

I also believe publishers when they say the economics of search feel different.

A preregistered field experiment from researchers at the University of Pennsylvania and Northeastern University gives us something better than vibes. The researchers recruited 1,100 U.S. adults, observed how they used Google, and then randomly assigned them to different search experiences. In the AI Mode-only condition, click-through to external publishers fell by 18.8 percentage points relative to the current Search experience.

That is not “AI killed 18.8% of traffic.” It is not a universal forecast. It was a specific experiment, over a specific period, using a specific version of Google’s product. The paper was also a preprint when I wrote this.

But the mechanism matters.

If the answer gets better inside the interface, fewer questions need a website visit to feel complete.

Search volume and referral traffic have split apart.

For a long time we treated those two things like they were basically the same curve. More searches created more chances to rank. More rankings created more clicks. More clicks created more opportunities to convert.

That chain still exists. It is just no longer guaranteed.

A person can now ask a complicated question, compare three options, add a follow-up, upload an image, ask for a summary, and narrow the answer down before they ever leave Google, ChatGPT, Perplexity or Gemini.

That means search can become more useful, more frequent and more conversational while the website receives a smaller percentage of the total interaction.

That sounds subtle.

It isn’t.

The answer is becoming the first landing page.

Traditional SEO assumed the search engine was a routing layer. The job was to help a person find the page where the real experience happened.

AI search turns the search engine into part of the experience itself.

The model retrieves information, assembles an answer, compares sources, and sometimes completes enough of the task that the user never needs the original page.

So websites now compete in two different ways.

They compete for the visit.

And they compete to become part of the answer before the visit.

This is why I think “zero-click” is too small a frame.

The phrase makes it sound like the only thing that happened was that somebody stole a click.

The bigger change is that discovery is being rebuilt around synthesis.

A machine can learn from your page, combine it with somebody else’s reporting, compare it against a third source, and then introduce your idea without ever showing your blue link as the main event.

That can be good for a brand. It can be terrible for a publisher. It can be both at the same time.

The measurement problem is that most of our dashboards were built for a world where the click was the handoff between discovery and understanding.

That world is already gone.

What I would measure now.

I would still measure rankings. I would still measure clicks. Anybody telling you SEO fundamentals suddenly stopped mattering is selling you something.

But I would add another layer.

Were we retrieved? Were we cited? Were we named? Which page was used? Did the answer describe us correctly? Did branded search rise afterward? Did direct traffic rise? Did leads mention an AI tool?

Those are not vanity metrics. They are the missing middle between “Google showed us” and “a person visited us.”

Google itself is moving in this direction. Search Console now has separate reporting for generative AI visibility in Search, including AI Overviews and AI Mode. That is a pretty strong signal that this layer is becoming measurable enough to matter.

The web is not disappearing. Its job description is changing.

I do not buy the dramatic version of this story where websites become useless and everybody lives inside one giant chatbot.

The models still need evidence.

They need original reporting, product details, firsthand experience, local knowledge, research, photographs, reviews, data, definitions, proof and people willing to put their name on a claim.

The web is still the source material.

But the source material does not always get the visit anymore.

So the real challenge is not “How do I get my clicks back?”

It is: How do I become valuable enough that both the machine and the human need me?

That is the game now.

Sources checked Sep 24, 2026

Google: A new era for AI Search

Google: New opportunities, control and insights for website owners

Alphabet Q2 2026: Search remarks

Wang et al.: AI in Search Reduces Publisher Referrals Without Improving User Experience