AI Search Reshapes Web Traffic as Publishers Push Back on Google

AI Search Grows, But Google Still Rules Referrals

New data from Similarweb suggests that AI search tools are not so much dethroning Google as building a new layer on top of it. According to the analysis, roughly 95% of ChatGPT's outbound clicks funnel to a small cluster of websites, meaning AI referral traffic is far more concentrated — and far less democratized — than the open web publishers built their businesses around 1. That concentration is forcing publishers to rethink what AI-driven visits are actually worth, since a handful of winners appear to capture nearly all of the referral benefit while the rest of the web sees little to none.

The Traffic Squeeze Publishers Are Feeling

The broader anxiety extends well beyond ChatGPT. Reports describe a growing "internet rebellion," with publishers blaming Google's AI Overviews for a measurable decline in the click-throughs that once sustained their ad and subscription revenue 6. Some of the biggest names in media — including Reddit, USA Today, Reuters, Politico, The Economist, and People Inc. — are reportedly weighing whether to restrict Google's access to their content altogether, a dramatic escalation that would have been unthinkable when search referrals were the lifeblood of digital publishing 7. Reddit itself posted a solid quarter financially, but executives and investors are increasingly nervous about the platform's dependence on Google and how an AI-reshaped web might disrupt that relationship going forward 3.

Commentary on Alphabet's strategy warns that the company may be undermining the very ecosystem that feeds its search engine. By surfacing AI-generated summaries that answer queries directly, Google reduces the incentive for users to click through to original sources, which critics argue is ultimately self-defeating: fewer visits mean less original content being produced for AI systems to summarize in the first place 4.

Measurement Problems Compound the Confusion

Adding to publishers' frustration is a data transparency issue. Google Search Console's reporting on AI Overviews reportedly shows impressions without corresponding clicks and artificially ranks every AI Overview link as position one, which distorts average position metrics and can mask real declines in traffic and revenue 5. That makes it harder for site owners to even diagnose how much damage AI summaries are doing, let alone respond strategically.

A Web Increasingly Built for Machines, Not People

Underlying all of this is a broader shift in who — or what — is actually browsing the internet. Some researchers now argue that AI agents and bots consume web content in fundamentally different ways than human readers, fueling renewed discussion of the once-fringe "dead internet" theory as automated traffic increasingly rivals or exceeds human activity online 2.

Taken together, the reporting paints a web in transition: AI tools are adding a new discovery layer atop traditional search, but the economics of that layer remain opaque, concentrated among a few winners, and increasingly contentious between Google and the publishers it has long depended on.