The number that made me stop and re-read it: 68.01% of US Google searches ended without a click between January and April 2026. That is Similarweb clickstream data across desktop and mobile, published by SparkToro.
A 7.56 percentage point shift in two years does not sound dramatic until you flip it around. The share of searches producing at least one click fell 9.51 points, which is a 22.9% relative decline. Roughly a quarter of the clicks that Google used to hand out, gone, in twenty-four months.
And that 68% figure includes clicks to Google's own properties. Maps, YouTube, the shopping panel. The share going to the open web, the part that lands on your site, is smaller than the leftover 32% implies.
But the number is not really the point of this post. The number has been written about to death, usually by people selling you a solution to it. The thing I have not seen covered honestly is the second-order problem, which is worse:
You cannot see this happening in your own analytics. By design. And the tooling that claims to fix that costs more than most indie products earn.
When someone searches, reads an AI Overview, and does not click, nothing happens on your server. No pageview. No referrer. No entry in any log you own. The event that cost you a visitor produced exactly zero data on your side.
What you see instead is a slow, unremarkable decline in sessions with no single cause you can point at. Traffic is down 8% quarter over quarter. Nothing broke. Your rankings are the same. Search Console shows your impressions holding steady or even climbing while clicks slide.
That last pattern, impressions flat or up while clicks fall, is the clearest fingerprint of AI-driven loss you have access to for free, and most people never look at it because they look at the clicks graph and not the ratio. If your position has not moved and your impressions have not moved but your CTR has dropped by a third, an answer above you is eating the click.
The scale of that effect is not subtle. When AI Overviews appear, CTR on the results below drops by nearly 60%. And Google's AI Mode, the fully conversational surface, produces a zero-click rate around 93%, more than double regular AI Overviews. If AI Mode becomes the default rather than a tab, the numbers in this post age badly in the wrong direction.
The zero-click statistic only measures Google. It does not measure the searches that never went to Google at all.
Somebody asks ChatGPT which tool they should use for X. The assistant names three products. Your product is either in that list or it is not. There is no impression, no ranking, no Search Console row. Nothing. If you are not named, the entire interaction is invisible to you, and it is invisible in a way that is qualitatively different from ranking eleventh, because at least ranking eleventh shows up in a report.
You get one weak signal: referral traffic. Visits with a referrer of chatgpt.com, perplexity.ai, claude.ai, copilot.microsoft.com. Those are worth segmenting out in your analytics right now if you have not, because that traffic converts unusually well. Somebody who arrives after an assistant recommended you by name is much further down the funnel than somebody who found you on a listicle.
But referral traffic only tells you about the times you won. It says nothing about the far larger number of times a competitor got named instead, which is the number you actually need.
I ran into this directly with roastme.gg. I put the site through ChatGPT's live search by hand to see whether it would show up for anyone not already looking for it by name. Doing that manually, one prompt at a time, is the crudest possible version of measurement, and it was still more informative than anything in my analytics. That is a bad sign for the analytics.
Before spending money, here is the honest list of what you can do for nothing. It is more than most people do and less than you need.
Segment assistant referrals. Build a report filtering referrers for the major assistant domains. Track it weekly. This is your "am I being cited" proxy, biased toward whichever assistants pass a referrer at all.
Watch the CTR ratio in Search Console, not the click count. Pull position, impressions, and clicks per query. Sort by queries where position is stable and CTR has fallen. That list is your AI Overview exposure.
Check your server logs for assistant crawlers. GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, and friends. Crawling is not citation, but never being crawled means citation is impossible. If they are not in your logs, check your robots.txt and your firewall rules before anything else, because plenty of people are blocking these by accident through a WAF preset.