Where did all the traffic go?

Here is a story I have now heard, in slightly different forms, from four different people in the last two months.

A site - a publisher, a niche shop, a how-to resource, it varies - has been doing everything right. Publishing regularly. Improving the content. The articles are better than they were two years ago: more thorough, better edited, more genuinely useful. Every internal quality metric is up.

And the search traffic is down. Not down a little. Down thirty, forty, in one case nearly sixty per cent year on year.

The person telling me the story is always slightly bewildered, because they’ve been told for fifteen years that good content earns traffic, and they have done the work, and the traffic has gone anyway. They want to know what they did wrong.

Mostly, they didn’t do anything wrong. The funnel changed shape underneath them.

The shape of the decline

If you go and look at the analytics for one of these sites, there’s a very specific signature, and once you’ve seen it a few times you recognise it instantly.

In Google Search Console: impressions flat or rising, clicks falling. Your pages are still being shown. Google still thinks they’re relevant. They’re appearing on results pages as often as ever - sometimes more. But the line for clicks has detached from the line for impressions and is heading down on its own. People in the SEO world have started calling this the great decoupling, and it is the single clearest fingerprint of what’s happening.

The reason is the one I wrote about in the disappearing results page. Your page got shown. An AI Overview at the top of the same results page summarised your page, along with four others. The user read the summary. Their question was answered. They didn’t need to click, so they didn’t. You earned the impression and lost the click, and the click was the only part that ever showed up in your traffic.

The asymmetry nobody agreed to

There’s a second number worth looking at, and it’s the one I find genuinely striking.

Cloudflare - which sits in front of a large fraction of the web - has been publishing what they call crawl-to-referral ratios: for every N pages an AI company’s crawler fetches from your site, how many visitors does that company send back to you?

For a traditional search engine the ratio was never one-to-one, but it was sane. Google crawled your pages and, in exchange, sent you a meaningful stream of visitors. That was the deal, even if nobody signed it: you let the crawler in, the crawler’s owner sent you traffic, the traffic was the payment.

The AI-crawler ratios are not sane. The figures Cloudflare has published put some AI crawlers in the range of hundreds, or thousands, of pages fetched per single visitor referred - and for a few, effectively no referrals at all. (Pull the latest numbers for whenever you’re reading this; the trend has only gone one way.)

Read that as an exchange and it’s extraordinary. The crawler takes your content - to train on, to summarise, to quote - and the old return payment, the traffic, mostly doesn’t come back. The page is still doing work. It’s informing answers. It’s just not informing them on your site, and the value of it is being realised somewhere you don’t have analytics for.

I want to be careful here, because this is contested ground. The AI companies would argue, not unreasonably, that being summarised and cited is itself worth something - brand exposure, authority, a citation chip with your name on it. That’s the argument from citations are the new ranking. It’s a real argument. But it’s also a different deal from the one most of these businesses built themselves on, and nobody sat them down and renegotiated it. It just happened.

The businesses that can’t absorb this

Plenty of organisations will be fine. If your relationship with a customer doesn’t depend on them finding you cold through a search engine - if you have a brand, an app, an email list, a sales team, a shelf in a shop - then a softer search funnel is a headwind, not a disaster.

The businesses that can’t absorb it are the ones whose entire model was the funnel. The independent recipe site funded by display ads on sessions that came from Google. The affiliate comparison site. The how-to publisher. The local-information site. For two decades the implicit contract was: produce indexable content, receive search traffic, monetise the traffic. Every link in that chain is now under tension at once.

And here’s the part that I keep coming back to, because it’s the actual point of this post: the people running those businesses can see the problem in their dashboards, and they do not yet have a vocabulary for it.

They know clicks are down. They can see the decoupling. But ask them "so what is your strategy for visibility in AI answers?" and - fairly - they don’t have one, because the question is barely two years old and the tooling to even measure the thing is immature. They’re still being sold "do more SEO", by people whose SEO advice was calibrated for a results page that no longer exists.

What comes next

So that’s the problem stated plainly. The traffic didn’t evaporate because the web got worse. It evaporated because the interface in front of the web changed, and the change quietly rewrote the deal that a large category of businesses depended on.

The obvious next question is: all right - what do you actually do about it? If being in the AI answer is the new being-on-page-one, how do you get there, and how do you know if you have?

That’s the next post, and it’s the one I’ve been building toward this whole series.