← All articles

Zero-click is the default now. What that does to your traffic model.

Most searches end without a visit. That isn't a traffic problem you can fix with more content — it's a measurement problem, and the teams handling it well have already changed what they report on.

There is a number that has been circulating for two years now, in various forms, and it has finally stopped being controversial: roughly six in ten searches end without anyone clicking through to a website. The exact figure moves depending on who is counting and what they count, but the direction has not wobbled once. Search increasingly resolves in place.

The usual reaction to this is to treat it as a content problem. If fewer people are clicking, publish more, publish better, chase the queries that still convert. That instinct is understandable and mostly wrong. The click was never the thing you wanted. It was a proxy for attention that happened to be easy to measure, and the proxy has broken while the attention has quietly moved somewhere your analytics cannot see.

The visit was never the point

Consider what actually happens when someone asks an assistant which project management tool suits a fifteen-person agency. They get a paragraph. It names three or four products, describes each in a sentence, and often adds a caveat about pricing. The person reads it, forms an opinion, and — if they go anywhere at all — goes directly to the product they have now decided to look at.

In your analytics, that shows up as direct traffic, or as a branded search two days later, or as nothing at all. The moment that mattered — the moment three names were considered and yours was or wasn't among them — left no trace on your side of the wire. You did not lose a click. You lost or won a shortlist placement, and you have no record of it either way.

What the traffic drop is really telling you

When organic sessions fall while revenue holds steady, the reflex in most companies is to assume a tracking error or a seasonal dip. Increasingly it is neither. It is the top of the funnel moving into a surface you do not own, while the bottom of the funnel — the people who already know your name — carries on behaving normally.

This produces a specific and dangerous pattern. Informational pages lose traffic first: the guides, the definitions, the comparison posts, everything a model can summarise in three sentences. Product and pricing pages hold up, because people who arrive there have already made a decision somewhere else. Aggregate traffic falls, conversion rate rises, and the dashboard looks like an efficiency win.

It is not an efficiency win. It is the visible part of your funnel shrinking while the invisible part grows. Give it four quarters and the pipeline follows, because the people who would have discovered you through those guides are now discovering three competitors through a paragraph instead.

Three things worth measuring instead

None of this means analytics is dead. It means the click-based metrics need company. In practice, three measurements cover most of what you lost.

The awkward part: these numbers wobble

Anyone selling you a clean, stable AI visibility figure is either simplifying or fibbing. Answer engines are generative. Ask the same question twice and you can get two different shortlists, with no change on your side and none on theirs. A single reading is a sample, not a rank.

The practical response is the same one you would use for any noisy measurement: repeat it, aggregate it, and report the trend with a confidence range attached. We run each prompt around thirty times a month per engine for exactly this reason. A day's movement means nothing. A fortnight's movement in the same direction means something.

This is also why you should be suspicious of any tool that reports a whole number with no error bar. It is not that the number is wrong. It is that a figure presented as precise, in a category this noisy, tells you the vendor has decided confidence is a better product than accuracy.

What to change this quarter

Start by separating your reporting into two halves. Keep the click-based metrics for the pages that still earn clicks — product, pricing, comparison, anything with commercial intent. Stop holding informational content to a sessions target it can no longer hit, and measure it on presence instead.

Then pick fifteen to twenty questions your buyers actually ask, in their words rather than your category's jargon, and establish a baseline. Not brand queries — those flatter you and tell you nothing. The unbranded, comparative, slightly awkward questions that come up on a first call.

Finally, accept that the fixes are slower than the measurement. Engines have to re-crawl, and models have to start drawing on what they find. Four to eight weeks is a realistic window for anything structural. The teams that do well with this are the ones who started the baseline early enough that they can prove the change when it arrives.