Google Zero: How the Web Stopped Paying for Itself

This was not a philosophical observation. It was a defensive argument in an antitrust case, deployed to suggest that breaking up Google's advertising business would only accelerate damage to publishers who already depended on it. When the filing drew attention, Google clarified that the phrase referred specifically to open-web display advertising, not the open web as a whole. The distinction is worth less than it sounds. A company that controls search arguing in court that open-web advertising is in terminal decline (to protect its own ad business from a breakup) is telling you exactly where it thinks traffic is going.

A few months earlier, Google's search SVP Nick Fox had told a podcast that "from our point of view, the web is thriving." The contradiction went largely unremarked.

At Google I/O 2026, the company announced what it called the biggest change to its search box in 25 years. Head of Search Elizabeth Reid described the result as "AI search through and through." The web Google admitted was in rapid decline is now being asked to sustain a system that no longer sends traffic back to it.

This is not a story about who is to blame. It is a story about a supply chain that consumed its own inputs, and what happens when those inputs run out.

The deal that ran the internet

For roughly twenty years, a transaction underpinned everything published on the open web. It was never written down, never negotiated, never acknowledged as a formal arrangement. But it was real, and almost every piece of content you have ever read online existed because of it.

Publishers produced, Google indexed, users clicked, and advertisers paid for the attention. The revenue funded the next article, the next investigation, the next recipe, the next how-to guide, and the cycle repeated as the web grew.

Search was not just a discovery mechanism. It was the economic infrastructure of an entire information layer. Google's crawler was, in effect, commissioning content by making it findable and therefore monetisable. The publisher did not need a relationship with Google. They just needed the traffic, and the traffic funded the work, and the work became the thing Google indexed to remain useful.

The system was extractive by design. Google took the value (attention, advertising yield, the data exhaust of billions of searches) and returned traffic as payment. Publishers called it organic reach. It was closer to a revenue share in which one party set the terms unilaterally and the other had no alternative.

The numbers

Zero-click searches (queries answered without the user visiting any external site) now account for roughly 60% of all Google queries, according to Similarweb and SparkToro. The rate climbs sharply with each step toward AI-native search.

Referrals are following. Google search referrals to publishers fell 33% globally in the year to November 2025 (Chartbeat, via the Reuters Institute); US publishers saw closer to 38%. These are aggregates. The distribution underneath is worse.

NPR described it as an "extinction-level event" for online news publishers. Condé Nast CEO Roger Lynch reportedly told colleagues to plan as if Google sends them no referrals at all. He does not expect it to reach literal zero, more likely a single-digit share of traffic, but for titles like The New Yorker and Vogue that is a distinction without much difference. The shorthand that has emerged in newsrooms, "Google Zero", looked hypothetical eighteen months ago. It now reads like operational planning.

When the Washington Post cut a third of its workforce in February 2026, Executive Editor Matt Murray was blunt about why.

The Post's sports desk, local coverage, foreign bureaux, and books section were reduced or closed. A newspaper built on comprehensive coverage is now concentrating on politics and national security, the areas where subscriptions survive the loss of search. This is not a marginal publication with a thin audience. This is the Washington Post.

What I/O 2026 actually announced

The industry tends to frame Google's AI search transition as something still approaching, something to prepare for. It has already happened.

At I/O 2026, Google did not announce a gradual feature addition. It announced a complete architectural replacement. The new Search is built around AI Mode, conversational follow-ups, and what Google calls "information agents": autonomous systems that monitor the web on your behalf and push updates to you over time. Every one of those features reduces the need to visit a source. The answer, the context, and the follow-up all arrive on Google's own surface, and the publisher never enters the picture.

The interface also builds custom visual experiences on the fly, pulling in structured data, images, and synthesised responses. This is no longer a summary sitting above a page of blue links. The links have become incidental, and the result page is itself the destination.

It is worth being precise about who benefits from this arrangement in the short term. Google does. Users staying on the results page rather than clicking through to publishers means more time on Google's own surfaces, more ad impressions on Google's own inventory. The company does not pay publishers for the content it summarises. Its revenues have continued to climb throughout the traffic collapse it is causing. The old deal sent traffic to publishers as a form of payment for content. The new deal sends nothing, while continuing to extract the same content. Google is currently profiting from a system it is simultaneously dismantling.

The regulatory response, and why it doesn't fix this

On 3 June 2026, the UK's Competition and Markets Authority issued a ruling described as a world first.

CMA chief executive Sarah Cardell called the measures a mechanism for "fair treatment" and "meaningful choice." Lawyers were more cautious. The framework sets a baseline, not a payment mechanism, and Google can slow-roll compliance through vague reporting obligations. The deeper problem is the shape of the choice itself.

The opt-out begins with a limited subset of UK publishers before any global rollout. The European Commission is watching under the Digital Markets Act, with a July 2026 deadline for Google to share anonymised search data with rivals. The pressure is genuine. It is also operating on a timeline that has nothing to do with publisher Q3 revenue. And the publishers most damaged by AI Overviews are usually the ones with the least leverage to exit: for a site already losing 60–70% of its search traffic, accepting still less visibility in exchange for content rights only accelerates the collapse it is meant to address.

The feedback loop nobody is talking about

Most coverage stops at the first-order effect: traffic down, revenue down, layoffs, closures. Correct, but incomplete. The second-order effect matters more, and it is almost absent from the discourse.

The content AI systems summarise was produced under the economics of the old deal. Publishers invested in reporting, research, and fact-checking because search traffic monetised that investment. Remove the incentive and the investment follows. None of this is speculative; it is already documented. The 2026 journalism layoff wave tracked by Press Gazette spans the Washington Post, Atlanta Journal-Constitution, Politico, Nexstar, Vox Media, Bustle Digital Group, CNBC, and the Wall Street Journal, all within the first two months of the year. Stereogum lost 70% of its ad revenue in 2025. The Planet D, a travel blog founded in 2008, lost roughly half its traffic after AI Overviews launched, then another 90% after further changes, and shut down entirely.

Put the commercial and the technical together and you get a loop that does not self-correct:

Google is not unaware of this. The legal filing admission about the open web's rapid decline suggests the company understands the trajectory. Understanding it and having a solution are different things.

What the adaptation strategies actually offer

The playbook recommended to publishers is consistent across the industry, and not wrong, exactly. But it is honest to be clear about what it addresses and what it does not.

None of these address what happens to the information layer when the organisations that cannot execute them (regional outlets, specialist publications, independent practitioners) stop being financially viable.

The citation volatility problem

There is a further problem that undermines even the "optimise for AI citation" response that has become common SEO advice.

The brands that do benefit, according to Amsive, are those with strong pre-existing brand entity signals: an 18% CTR lift on branded queries under AI Overviews. But you do not build brand entity authority in response to AI Overviews. You either have it or you do not. For most publishers, the window for building it under the economics of search traffic has already closed.

What this is, structurally

The standard framing is that Google disrupted publishing. The more accurate framing is that Google internalised it.

Google did not set out to destroy publishing. It set out to answer queries more effectively. The damage is a side consequence of a product decision, which is, if anything, a worse prognosis than deliberate harm. Deliberate damage can be regulated, negotiated, or litigated. Side effects of a popular product improvement are far harder to address: they are not the intent, they are hard to attribute, and the product causing them keeps getting more useful to users.

The users are getting better answers. The system producing those answers is consuming the inputs that made good answers possible. Both are true at once, and they point in opposite directions.

The open question

The adaptation strategies (owned audience, first-party data, diversified revenue, licensing) work for some publishers under some conditions. They are not a systemic answer to what happens when the economic model for open-web publishing collapses.

The honest framing is not pessimism. It is structural logic. Content that costs money to produce requires a funding mechanism. When the primary mechanism fails, less content gets produced. When less gets produced, the information layer thins. When it thins, the AI trained on it has less to work with. The summary quality of a system trained on a depleted open web will not match one trained on a fully-funded one.

What replaces the search-traffic subsidy? Subscriptions, for brands strong enough to command them. Licensing deals, at rates set by the AI companies. Public funding, in markets that support public media (already under pressure in the US, where NPR is restructuring after losing federal subsidies). Philanthropy. Creator economies built on individual trust rather than institutions. These are not nothing. They are also not equivalent to the infrastructure that funded the open web for twenty years.

The Goodhart's Law piece on this site argued that Google has been playing a structurally unwinnable game against signal gaming for two decades. This is the conclusion of a related dynamic: an extractive relationship that sustained the content layer just long enough to consume it.

"Google Zero" is not a worst case modelled by risk-averse executives. Condé Nast's CEO told his teams to plan for it. It is already the operating reality at NPR, at regional newsrooms without brand leverage, at specialist sites with no acquisition channel beyond search.

The web that trained the AI is not the web that will be left after the AI has finished consuming it. The question nobody has answered yet is what we do with the difference.