Opinion. The measurement behind this is in the zero-click study. The argument is mine, and you may disagree with it.

On 18 November 2026 I searched 40 everyday UK questions on Google. Every results page put an answer above the first ordinary link. 38 of them were AI Overviews. On 22 of the 40, no ordinary link was visible without scrolling.

That is one morning, one location and desktop only. It measures what Google shows. It says nothing about what people click, so I won’t claim traffic has fallen by any amount. I don’t have that number.

What it does let me say is this: for a simple question, the answer now comes first and the link second. So what is the link for?

The job that has gone

For years a page could earn a visit by holding a simple fact. A date. A rate. How long something takes. If my reading holds up when it is repeated, that job is finished. Google does it on the results page, before anyone reaches your link.

I don’t think that job is coming back. I also don’t think it was ever a safe one. A page that exists to repeat a fact someone else owns was always one product change away from this.

What is left

Two jobs survive, in my view, and I have some evidence for the first.

The first is being the source. On 18 of the 40 questions the first ordinary link went to a public body: gov.uk, the NHS, the Bank of England or MoneyHelper. They hold that spot because the fact is theirs. An answer box has to get its fact from somewhere. If the number, the date or the dataset started with you, you are much harder to leave out.

The second is doing the thing. An answer can tell you what something costs or when it is due. It can’t log you in, show you your own account or take your order. Those still need somewhere to go.

There may be a third: being a site people ask for by name. I can’t measure that yet, so treat it as a hunch.

Where publishers are

The UK’s ten biggest newsbrands blocked 99 of 150 site and AI crawler pairs in their robots.txt in October, and the same 99 in November. Nothing moved.

That is a different fight from Google’s results page. Blocking GPTBot does nothing to an AI Overview. But it shows the instinct, which is to protect the page.

I think the page is the wrong thing to protect. The fact, the dataset and the name are worth more than the page they sit on.

What I’d do with a site today

  • Publish something that starts with you: a count, a price list, a dataset. Put the date on it.
  • Make the thing people come to do easy to reach.
  • Stop writing pages whose only job is to restate a fact another site owns.

What would change my mind

The same 40 questions get read again on 18 December 2026 and every month after that. If fewer of them show an answer above the first link, or the first link moves back up the page, I’m wrong about the direction and I’ll say so here.