When Google answers the question on the results page, some of the clicks that used to reach your site never happen.
That's not a hot take. It's the visible mechanic of AI Overviews: one assembled answer, a few cited sources, and a lot of queries resolved without anyone leaving Google.
The doom version of this story ends there. Ours doesn't, because "fewer clicks" and "nothing left to win" are different claims, and only the first one is true.
The part that's genuinely gone
Some traffic isn't coming back.
The informational query that wanted a quick fact, got it in the Overview, and closed the tab - that click is gone, and no amount of optimisation retrieves it. Content strategies built entirely on high-volume informational traffic are the casualty here, and pretending otherwise wastes everyone's time.
Grieve briefly. Then look at what the new format actually rewards.
The part that's still winnable
The answer has to come from somewhere. AI Overviews cite sources. Assistants cite sources. Those citation slots are the new contested ground, and they're being claimed while a lot of businesses are still arguing about whether any of this matters.
Being the cited source gets you two things. The clicks that do still happen - from people whose question wasn't fully resolved and who want the source. And presence inside the answer itself: your name attached to the recommendation, even when no click follows.
That second one feels intangible until you notice it's the same mechanism as being recommended by ChatGPT for a buying question. One answer, a few citations, and either you're in it or your competitor is.
What earns the slot is what we build in our AI search optimisation work: crawlable, well-structured pages, consistent entity signals, content that makes clear claims and answers questions directly so it can be lifted accurately, and third-party corroboration backing up what you say about yourself. We've covered the structured-data side of this in detail.
None of it guarantees placement - nothing does, and be suspicious of anyone who says otherwise. It improves every factor these systems demonstrably weigh. That's the honest ceiling of the discipline right now.
What to measure instead of mourning
If clicks are a shrinking window onto your visibility, the measurement has to widen. Ours looks like this.
Structured prompt testing across the major assistants: do you appear for your buyers' questions, how are you described, who's cited instead. Tracked over time, because single runs mean nothing.
AI referral traffic in analytics - what's arriving from assistants and Overviews, and what it does after landing. The volumes are often small. They're also often further along than the average organic visit, and worth watching for that reason.
And your own search data, read for the new pattern: impressions holding up while clicks thin out isn't a rankings problem, it's the format changing shape around you.
It's early-days measurement. We say so to clients, and we'd rather have honest, limited numbers than confident fiction.
The reframe
The question used to be "how much traffic do we get". Increasingly it's "when a machine answers our buyer's question, are we in the answer".
Fewer clicks, higher stakes per appearance. That's the trade.
The businesses that adjust to it early will take the citation slots while their competitors are still reporting on rankings.
