The metric that defined digital performance for two decades is quietly losing its grip.
Across all Google searches, 43% already end without a click. In AI Mode, that number rises to 93%. It’s a striking shift—but not for the reason most teams assume.
“The 93% figure is the cleanest signal we have that the click was never the thing we were actually measuring,” Shane H. Tepper, cofounder of Resonate Labs, a company that helps B2B businesses be found and cited in AI search models like ChatGPT, Perplexity and Gemini.
For years, the click functioned as a proxy. For interest, for intent, for progress toward a decision. In AI-driven interfaces, that proxy disappears. The decision it represented does not.
“Google answers the question in place and the user moves on, so the proxy disappears. The decision it stood for didn’t.”
That distinction reframes the entire conversation. What looks like a traffic decline is, in reality, a relocation of behavior. Research hasn’t slowed down. It has moved upstream, into environments that traditional analytics can’t see.
“Demand formation moved upstream of the click, into a layer your analytics can’t see,” Tepper says.
The implication is uncomfortable. Most performance systems are still built around measuring what happens after a user arrives. But in AI-mediated discovery, much of the evaluation process is completed before a visit ever occurs.
And in many cases, that visit never comes at all.
The Attribution Illusion
Even when AI-driven journeys do result in traffic, the underlying influence is difficult to detect.
“Ahrefs confirmed in January 2026 that clicks from Google AI Overviews land in analytics as ordinary Google/organic traffic, with no native way to isolate them in Search Console or GA4,” Tepper explains.
The more common scenario is harder to trace. A buyer spends time inside an AI interface comparing vendors, closes the session, and returns days later via a direct visit or branded search. The analytics system records a familiar entry point. The interaction that shaped the decision is missing.
“The influence gets laundered into channels that already look familiar.”
This creates a structural blind spot. Teams continue optimizing for channels they can measure, while a growing share of decision-making happens elsewhere. Performance appears stable. Attribution appears intact. The underlying dynamics have already shifted.
That’s why conversion data alone can be misleading.
“Conversion rate is not attribution, and treating it like one is how teams talk themselves into believing they understand a channel they can’t see.”
AI-influenced traffic often converts at a premium, particularly in B2B. But higher conversion rates don’t explain how demand was created, only that it arrived more qualified.
To understand the source of that qualification, teams have to look beyond the click.
From Traffic to Presence
If clicks are no longer a reliable proxy for intent, the question becomes: what replaces them?
“The replacement is a small set of measures that describe presence in the answer layer instead of traffic to your site,” Tepper says.
That layer operates on a different logic. Instead of ranking pages, AI systems retrieve and synthesize information into responses. Visibility is determined not by position, but by inclusion, whether a brand appears in the answer at all, and how it is represented when it does.
Four signals define that presence.
First, presence itself: how often a brand appears across the queries that matter in its category. Not in isolated prompts, but across repeated runs and real research patterns.
Second, citation share: how frequently a brand is named relative to competitors within those responses. Showing up is not the same as leading the conversation.
Third, narrative accuracy: what the model actually says about the brand, and whether that description reflects reality.
“A brand that appears consistently but gets described with outdated pricing or a competitor-planted limitation has a visibility problem dressed up as a visibility win.”
And fourth, platform-specific visibility. Each AI system retrieves from a different pool of sources, with minimal overlap. Aggregating performance into a single “AI visibility” metric obscures more than it reveals.
“A single aggregate score averages four different channels into a number that hides all of them.”
Reading the Signals You Can’t Track
In the absence of clean attribution, teams are left with indirect indicators, patterns that reveal influence without explicitly measuring it.
“When your AI citation share climbs, branded and direct sessions usually follow one to three weeks later,” Tepper notes.
The lag reflects real buyer behavior. Research happens inside the AI interface. Follow-up actions appear later in analytics. The connection is visible only if teams know where to look.
Other signals emerge in less structured ways. Language from AI responses begins to surface in sales conversations. Leads arrive with clearer preferences, asking validation questions instead of exploratory ones.
“None of these is clean attribution. They’re fingerprints, not footprints.” But in a system where direct attribution is structurally unavailable, those fingerprints offer the closest approximation of how decisions are being shaped.
A Different Optimization Target
The shift to AI-mediated discovery doesn’t eliminate the need for visibility. It redefines what visibility means. “If you’re not in the interaction, you’re not in the decision.”
In a world where most searches no longer produce a click, the primary competition is no longer for traffic. It’s for inclusion in the answer itself.
And that changes what there is to optimize.