A reporting problem has emerged that looks like a data error and is not: search impressions rising while clicks fall.
Why it happens
Zero-click searches reached 68% in early 2026, up from 60% in 2024. When an AI Overview appears, click-through at the top organic position drops 58% — up from 34.5% in April 2025.
A page can therefore appear more often, in a better position, and receive fewer visits. Impressions measure appearance; clicks measure onward journeys. Those have decoupled.
What breaks in reporting
Position as a proxy for traffic. Improving from position 4 to position 2 no longer implies the traffic gain it did in 2023.
Year-over-year comparison. The same query, same position, same impressions can produce materially different clicks in 2024 and 2026. The comparison looks valid and is not.
Channel attribution. With authenticated match rates near 47% against roughly 68% in the cookie era, unresolved conversions land in direct or organic buckets rather than being attributed to the touchpoints that influenced them.
Two independent degradations, both in the direction of under-crediting paid work.
How to report honestly
Show impressions and clicks separately, never blended. A blended visibility score hides exactly the divergence you need to see.
Segment by query type. Informational queries lose clicks fastest; transactional and comparison queries hold up better. Aggregate CTR mixes two different trends into one misleading number.
Track CTR as its own metric over time. A declining site CTR at stable positions is the fingerprint of this shift, and it is invisible if you only look at sessions.
Use holdouts for incrementality. Geographic or audience holdouts measure causation without depending on identity resolution or click attribution. They are the only method that did not degrade.
Scale of the effect
Google Search referrals across more than 2,500 news sites fell 34% year over year, with small publishers down 60%, mid-sized down 47% and large down 22%.
And AI chatbots, despite growing referral volume more than 200%, remain under 1% of total publisher referrals — so the traffic is not simply moving somewhere else you could measure instead.
The reporting job now is to describe a market where visibility and traffic are separate things. Dashboards built when they were the same thing will keep producing confident, wrong answers.
