Your Traffic Dropped and Your Rankings Did Not. Here Is How to Tell Which Problem You Actually Have.

Somewhere in the last year, a strange support conversation became common across the industry. A business owner notices traffic is down, checks their rankings, and finds them fine. Still position three for the terms that matter. Fewer visitors every month anyway.

The usual suspects get blamed in order. The developer. The hosting. A penalty. A redesign. Sometimes money gets spent fixing things that were never broken.

What is actually happening, in a growing number of these cases, is simpler and stranger: the site did not lose its ranking. The ranking stopped producing clicks.

## The numbers, with an honest caveat

In June, Rand Fishkin’s SparkToro published a study using Similarweb clickstream data covering the first four months of 2026. The headline finding: 68.01 percent of US Google searches ended without a click on anything. Not your site, not a competitor, not a news article. The search happened, an answer appeared, and the person left.

Two years earlier, the comparable figure was around 60 percent. A decade ago, by older measurements, it was under half.

Separately, Pew Research tracked real user behavior across tens of thousands of actual searches and found that when an AI generated summary appears at the top of the results, people click a traditional result only 8 percent of the time, versus 15 percent when no summary is present. Clicks on links inside the summary itself: about 1 percent.

Now the caveat, because it matters and almost nobody covering this includes it. These studies use different panels, different years use different data providers, and the derived statistics being passed around social media conflict with each other. We found at least three incompatible versions of “clicks per thousand searches reaching the open web” in circulation, all citing the same study. Fishkin himself describes the year over year comparison as apples and oranges. The direction is reliable. The precise slope is not. Anyone quoting these numbers to two decimal places as proof of anything specific is selling something.

What the numbers support saying is this: a meaningfully growing share of searches end without sending anyone anywhere, the presence of an AI summary roughly halves the odds of a click, and none of this shows up as a ranking change.

## Two problems that look identical from the dashboard

This is the part that is useful on a Tuesday.

When traffic falls, there are two structurally different causes that get treated as one.

The first is a ranking problem. Google moved you down, or moved someone else up. Your visibility genuinely decreased.

The second is a click problem. Your visibility is unchanged, but the page of results you appear on now answers the question before anyone scrolls to you.

The fixes for these two problems have almost nothing in common. Treating a click problem as a ranking problem means paying for SEO work to defend a position you never lost. Treating a ranking problem as a click problem means shrugging at something that is actually fixable.

## How to tell them apart in fifteen minutes

Search Console distinguishes them, though it does not label the distinction.

Open the Performance report and look at three curves together over the last twelve months: impressions, average position, and click through rate.

A ranking problem looks like position falling and impressions falling together, with CTR roughly stable. You are being shown less often because you sit lower.

A click problem looks like position stable and impressions stable or even rising, while CTR falls. You are being shown just as often. Fewer of the people seeing you have a reason to arrive.

Check it query by query for the searches that matter to your business, not as a site wide average, because averages blend the two problems into fog. A site can have a ranking problem on one set of terms and a click problem on another at the same time.

If what you find is the second pattern, your SEO did not fail. The deal changed.

## What actually follows from a click problem

There is no setting that fixes this, which is uncomfortable, but the diagnosis still changes decisions.

Informational queries are affected far more than transactional ones. Searches like what is DNS propagation get answered in place. Searches like emergency plumber near me or the name of your business still produce clicks at close to normal rates, because the person needs a destination, not an answer. The practical consequence: content built purely to catch informational search volume is depreciating, while being the business people search for by name is appreciating.

Which points at the response that holds up over time. The measurable asset is no longer how many strangers arrive from generic questions. It is how many people search for you specifically, arrive directly, or come back on purpose. Branded search volume, direct traffic, an email list you own. Those are the numbers that cannot be taken away by a change in how a results page is drawn.

None of that is fast, and we would be skeptical of anyone claiming a fast version exists.

## The pattern this belongs to

This is the same story as the Cloudflare default change we wrote about this week, from the other side. Decisions about how much traffic the open web receives are increasingly being made by the platforms that sit between you and the people searching, they arrive as defaults and interface changes rather than announcements, and they never show up in the metrics you were told to watch.

The useful response is not alarm. It is knowing which problem you have before paying to fix the other one.

*Sources: SparkToro/Similarweb zero click study, June 2026. Pew Research Center analysis of user click behavior with AI summaries.*

—

**About Blue Arctic** — Blue Arctic has provided managed hosting since 2002, on infrastructure we own in Tampa, Florida. Questions about what your own traffic numbers are telling you are always welcome.