Are people really leaving Google for AI search—or are they simply changing how they use it?

The answer may be far more complicated than the headlines suggest.

In this episode of The Deep Dive, we unpack one of the biggest contradictions reshaping digital marketing and the open web: reports suggest that most ChatGPT users still use Google, yet publishers, creators, developers and businesses are seeing meaningful declines in search referrals and outbound clicks.

So, is Google still winning—or are the metrics measuring the wrong thing?

We examine the difference between audience reach, search frequency and actual clicks, and why the widely cited 95% overlap between Google and ChatGPT users may not tell marketers what they think it does.

The discussion explores recent research and clickstream evidence, including findings associated with Similarweb, Comscore, Bocconi University and experimental research into Google’s AI-powered search experience.

One of the most important findings discussed is the distinction between being a Google user and actively relying on Google for information discovery. A person who performs a single Google search in a month can still be counted as a Google user—even if most of their complex research, writing, coding, brainstorming and information discovery happens through an AI platform.

That distinction changes the story.

From search engine to answer engine

Traditional search sends users outward. AI-powered search increasingly gives users the answer directly.

This shift has major implications for publishers, news organizations, academic websites, reference sites, developer communities and content creators that depend on organic search traffic.

The episode explores reported declines in referrals to several informational categories, including:

• Academic websites
• Reference and encyclopedia websites
• Developer and coding resources
• News websites
• Community-driven platforms

At the same time, transactional destinations such as marketplaces and entertainment platforms appear more resilient.

The result is a potentially fundamental change in the architecture of the internet: AI may increasingly become the destination for information discovery, while traditional search becomes more of a utility for navigation, commerce and specific destinations.

The rise of AI visibility and the “dark funnel”

The transformation doesn’t stop at declining clicks.

A new attribution problem is emerging.

A consumer might discover a brand through ChatGPT, Gemini or another AI assistant, then later search for that brand on Google and visit its website.

Analytics may attribute that visit to Google organic search.

But Google did not create the original demand.

The AI did.

This invisible journey represents what we call the AI dark funnel—a growing gap between where influence happens and where the final measurable click occurs.

That means marketers may need to move beyond traditional SEO metrics and begin measuring AI visibility, brand mentions, AI recommendations and presence within generative answers.

Is AI killing the open web?

Perhaps the biggest question isn’t whether people are abandoning Google.

It’s whether the internet can sustain itself when AI systems increasingly consume information from the open web but send fewer users back to the publishers and communities that create it.

If informational websites lose traffic, advertising revenue and subscribers, what happens to the incentive to produce high-quality information?

And if those creators stop publishing, what information will future AI systems have left to discover, synthesize and cite?

This episode explores that uncomfortable feedback loop—and why the future of search may ultimately depend on finding a sustainable relationship between AI platforms, search engines, publishers, creators and the open web.

The real question isn’t “Are people leaving Google?”

It’s “What are people still using Google for—and where does their attention go before they get there?”