Google's AI search summaries are redirecting web traffic away from publishers

AI-Generated Search Summaries Are Quietly Reshaping Internet Traffic

Google's parent company Alphabet posted a 24% year-over-year revenue increase last quarter, with much of that growth fueled by AI infrastructure and services. Beneath that impressive figure, however, a structural shift in how people use Google is beginning to redistribute — and in many cases eliminate — the web traffic that publishers, small businesses, and digital media organizations have depended on for years. The culprit is Google's AI Mode, which now places AI-generated summaries at the top of search results, often answering a user's question before they ever reach an external website.

The scale of this change has been rapid. An outside study found that AI summaries appeared in roughly 15% of Google searches about a year ago; today that figure has climbed to approximately 40%, and the feature is now effectively a global default. That shift has profound consequences for the economics of online publishing. When a user gets an answer from a summary rather than clicking through to a source, the publisher loses a pageview — and with it, advertising revenue and the potential for a longer engagement that might lead to a subscription.

Publishers Are Bleeding Traffic, and the Numbers Are Stark

The damage is already measurable. Business Insider has reported a 35% year-over-year decline in web traffic. The Huffington Post and The Washington Post have documented similarly steep drops over the past several years, a period that coincides with the mainstreaming of generative AI tools in search. Even Wikipedia — which operates as a nonprofit and exists specifically to make information freely accessible — has seen human visitor traffic fall roughly 8% over the last year alone.

Google has pushed back on the most alarming characterizations of this trend. In a blog post, the company argued that it continues to send billions of users to external sites and that outside research exaggerates the problem. One study estimated that 75% of users do not click through to any website when an AI summary is available — a figure Google disputes as overly pessimistic compared to its own internal data. The company also points out that AI summaries include links, meaning discovery of external sites has not disappeared entirely.

That defense, however, does little to reassure publishers watching their analytics. The disagreement between Google's internal figures and independent research reflects a fundamental information asymmetry: the company that controls the search experience also controls the most complete data about how that experience affects outbound traffic, creating an inherent conflict of interest in how those numbers are framed and released.

A Two-Tier Internet: Big Players Cut Deals, Everyone Else Struggles

The competitive landscape is fracturing along lines of scale. Large, well-known media brands with significant subscriber bases have been able to negotiate their way into Google's AI summaries. Google has built publisher profiles that, for subscribers of a given outlet, surface that outlet's content within AI-generated results. Beyond that, major institutions including The New York Times, the Associated Press, and Reuters have struck licensing deals directly with AI companies, ensuring their content is used — and compensated — in the training and deployment of AI systems.

For smaller publishers, independent journalists, niche websites, and local news organizations, neither of those paths is readily available. Without the negotiating leverage that comes from millions of subscribers or a globally recognized brand, these outlets are unlikely to land meaningful licensing deals or earn prominent placement in AI summaries. Their content may still be scraped and used to generate those summaries, but the traffic and ad revenue that once followed are no longer guaranteed to flow back. Discovery — the organic process by which a user stumbles onto a new site through a search result — becomes increasingly rare when a summary resolves the query without a click.

This dynamic threatens to accelerate a consolidation of the media ecosystem that was already underway before AI search arrived. The outlets with the resources to adapt will likely survive; those without may find that their business model — built on search-driven page views and display advertising — has been quietly dismantled.

Cloudflare's Proposal and the Future of an Open Web

One emerging response to this crisis comes from Cloudflare, a major internet infrastructure company whose services underpin a significant portion of global web traffic. Cloudflare has proposed shifting the monetization model away from human click-throughs entirely, instead charging AI companies and their bots for crawling and accessing content on publisher websites. Rather than earning revenue when a person visits a page, a publisher would earn revenue when an automated system reads that page in order to generate an AI summary.

The idea has intuitive appeal — if AI systems extract value from content, those systems should pay for it — but it also implies a profound change in what the internet is and who it serves. A web in which the primary transaction is between publishers and AI crawlers, rather than between publishers and human readers, is structurally different from the open, user-facing internet that has existed for three decades. It risks creating a more siloed, commercially negotiated information environment, where access to content is mediated by licensing agreements rather than a shared digital commons.

The broader stakes extend beyond any single industry. The internet as a discovery mechanism — the idea that anyone with a website could reach a global audience through search — is increasingly contingent on decisions made by a handful of AI companies and one dominant search engine. Whether that model proves durable, or gives way to a more fragmented and monetized alternative, will shape how information is produced, distributed, and funded for years to come.