We Stopped Clicking

A while ago I sat with a friend, trying to work out why his website was bleeding visitors. Nothing was broken. The page still ranked high on Google, like it had for years. Yet the traffic kept sliding, month after month, like someone slowly turning down the volume.

The explanation was higher up the screen. Above his link, Google had started printing the answer itself — summarized, finished, pulled partly from his own text. The visitor got what they came for. They never had to click.

That's not a bug in his site. It's the web changing shape.

From place to interface

For twenty years the web was a place you visited. You searched, you got ten blue links, you clicked through. The whole economy of the open web — ads, subscriptions, reach — rested on that click.

The click is disappearing.

In early 2026, roughly two in three Google searches ended with no click at all — the user got their answer without leaving the page. When Google shows its AI summary up top, click-through falls by around 60%, and for those searches up to 80% end in zero clicks. The exact numbers vary between studies. The direction doesn't: fewer and fewer questions ever reach a website.

And it's still early. Google's dedicated "AI Mode" is still a sliver of all searches — but its users roughly double every quarter. Add ChatGPT, Perplexity and the rest, and the pattern gets hard to unsee: more and more often, it isn't you reading the web. It's an AI reading it for you and handing you the answer.

The web stops being a destination. It becomes substrate — something machines read on your behalf. Raw material. Bot-food.

What dies, and what survives

"SEO is dead" is the wrong conclusion — and an overclaim a sharp reader sees through instantly. SEO isn't dying. It's changing the game.

What actually dies is the brochure site: the informational page that lived on search traffic, on being the answer someone googled and clicked to find. When the answer is delivered straight into the interface, there's no reason to visit. That traffic isn't coming back.

Three things survive:

Surfaces that convert. Once an AI does send someone, it's because they have intent — to buy, to book, to decide. Ten blue links don't matter then; only whether the page makes it happen.
Products and apps. Surfaces where you do something, not just read something. An AI can summarize a text for you. It can't be the tool for you.
Being the source AI cites. The new game is called GEO/AEO — getting pulled and referenced by the answer engines, not ranking among the blue links. Gartner expects traditional search volume to fall ~25% by 2026. Attention isn't leaving. It's moving up — into the answer.

The real question nobody's asking

Here's the part I can't let go of.

When AI becomes the interface between people and everything they want to know, it all boils down to one question: which thing gets shown to this person? Which answer, which product, which source, which supplier — out of a thousand possible — the AI surfaces.

That's not a search problem anymore. It's a matching problem. And right now it's being solved the worst way we know: by ranking. A handful of answer engines sit between all supply and all demand and rank the world for us — on their criteria, in their order.

We've seen what one-sided ranking does. Everyone measured against the same yardstick, everyone chasing the same top few. The winner takes almost everything, the rest goes invisible. The open web's great promise was distribution — anyone could reach out. The answer layer re-centralizes it into a few gates, with exactly the concentration ranking always produces.

And ranking is not matching. Ranking asks "what's best?" and gives everyone the same answer. Matching asks "what's right for you, given everything and everyone else at once?" — and gives different answers to different people, so the whole thing fits together. One concentrates. The other distributes. We're building the most important discovery layer in the world on the wrong one of the two.

What it means to build now

If you're building a site to be found in search, you're building for a world that's already closing. If you're building a surface that converts intent, a product that actually does something, or a source worth citing — you're building for what's coming.

But the bigger insight sits one level up. When discovery becomes matching, the question isn't how do I rank higher. It's who decides what gets shown — and do they rank, or do they match? Because that decides whether the new web is just another winner-take-all place, or something that genuinely pairs the right thing with the right person.

The people building the old way are on the wrong side of a shift you can see coming. Not because they're doing something foolish — but because the ground moved under them.

So the question isn't whether AI is eating the web. It already is. The question is what takes its place:

A layer that ranks the world for us — or one that matches it?

Tim Brandin has spent eight years building the matching kind — it's most of what he thinks about at tr8s.