The Zero-Click Internet: When AI Becomes the Only Door to Information

Something strange is happening to the internet. The fundamental transaction that has powered the web for three decades (you search, you click, you visit a website) is breaking down. And most people haven’t even noticed.

In 2024, roughly 58% of Google searches ended without a single click to any external website¹². By mid-2025, that number had climbed past 69%³. For news searches specifically, based on one report, click-through rates dropped by as much as 89% when AI Overviews appeared for their content⁴. The information flows, but nobody visits the source.

Welcome to the zero-click internet.

What Zero-Click Actually Means

The traditional web worked on a simple premise: websites create content, search engines index it, users find it through search, and traffic flows back to the source. Advertising revenue follows eyeballs. Publishers get paid. The cycle continues.

AI has inserted itself into this cycle in a way that fundamentally disrupts it.

When you ask ChatGPT a question, it synthesizes an answer from its training data, data scraped from the very websites you’re now not visiting. When Google shows an AI Overview at the top of search results, it answers your query directly, making the links below feel almost… optional. When Perplexity summarizes a research topic for you, complete with citations, the citations become footnotes rather than destinations.

The information still exists. You still get your answer. But the websites that created, researched, verified, and published that information? They become invisible infrastructure. Used but unvisited, essential but economically starved.

The Numbers Are Brutal

The numbers should concern anyone who cares about the health of information on the web.

When Google’s AI Overviews appear on a search result page, click-through rates to organic results drop. For paid ads, the decline is even steeper. Users are nearly 34.5% less likely to click on anything when an AI summary answers their question⁵.

The impact on publishers has been catastrophic. Business Insider’s organic search traffic fell by 55% between 2022 and 2025 ⁶. HuffPost lost half its organic search traffic ⁶. Stereogum, a music blog, saw 70% of its ad revenue evaporate. CNN experienced a 30% traffic decline.

On top of that, AI search tools like ChatGPT and Perplexity, despite their explosive growth, aren’t sending meaningful traffic back. They drive roughly 95-96% less referral traffic than traditional Google search. Click-through rates from AI answers hover below 1%.

The math doesn’t work. AI companies consume the output of publishers to train their models and generate answers. Users get what they need without visiting. Publishers lose the traffic that funds their operations. The well is being drained faster than it’s being replenished.

The Agentic Future Makes This Worse

What I’ve described so far is already happening. But we’re standing at the edge of something even more transformative.

2025 has been the year of AI agents, systems that don’t just answer questions but actively browse the web, fill out forms, make purchases, and complete tasks on your behalf. OpenAI launched ChatGPT Agent. Google is testing Project Mariner. Microsoft is building autonomous browsing into Edge’s Copilot. Startups like Perplexity, Arc, and others are reimagining the browser as an AI-first interface.

These aren’t just search replacements. They’re browsing replacements.

Imagine asking your AI assistant to research the best laptop for your needs, compare prices, read reviews, and make a purchase. The agent visits dozens of websites, extracts information, and synthesizes it, but you never see any of those pages. You never experience the publisher’s content, their advertising, their subscription prompts. The agent does the browsing; you just get the outcome.

At that point, the question isn’t just “will people visit websites?” It’s “will people browse at all?”

The Curations Problem

I’ve written before about the asymmetry problem in AI trust, and how our increasingly intimate relationships with AI create dangerous power imbalances. The zero-click internet is a direct manifestation of this concern.

In that piece, I mentioned:

As we head toward a zero-click internet, with these AI models soon becoming our primary ways of obtaining information and making decisions, the “curations” done by these models and with what larger objectives they are done become crucial in how we perceive our world.

This is exactly what’s unfolding. When AI becomes the primary interface to information, whoever controls the AI controls the curation. And curation, as we’ve learned from the social media era, is never neutral.

Search engines already made curatorial choices: which results to show, in what order, with what emphasis. But you could at least see those results and make your own judgment. You visited the source and formed your own impressions from the original content.

AI answers collapse that process. The model synthesizes, summarizes, and delivers a finished response. You don’t see the twenty articles it drew from. You don’t encounter the dissenting view it weighted less heavily. You don’t notice the nuance that got lost in summarization. The curation becomes invisible.

The Verification Problem

Something that bothers me deeply: verification becomes nearly impossible in a zero-click paradigm.

When you read an article on a news website, you have context clues for evaluation. Who published this? What’s their track record? Does the writing seem professional? Are claims sourced? What’s the publication date? These signals, imperfect as they are, help you calibrate trust.

AI answers strip away this context. You get information without provenance. Facts without context, conclusions without the reasoning that led to them.

Yes, some AI systems include citations. But citation is not the same as verification. A link at the end of a paragraph doesn’t mean you’ll click it. And even if you did, you’d be visiting the source after you’ve already absorbed and possibly internalized the AI’s interpretation.

The direction of trust inverts. Instead of evaluating sources to form conclusions, you receive conclusions and optionally verify sources. It’s the difference between building understanding from the ground up versus accepting a pre-constructed narrative.

The Economic Collapse of Information

So what happens when the advertising model fails?

Digital publishing has been precarious for years. But it functioned. Websites created content, attracted traffic, sold ads against pageviews, and used that revenue to fund more content creation. It wasn’t perfect, and clickbait and engagement farming were obvious downsides, but it sustained journalism, analysis, research, tutorials, and countless other forms of valuable information.

The zero-click internet undermines this model at its foundation.

Industry estimates suggest that AI-powered search summaries reduce publisher traffic by 20-60% on average, with niche publications experiencing losses increasing. That translates to millions in annual advertising revenue losses across the publishing sector. Forrester predicted that brands would cut open web display spending by 30% in response to AI search adoption⁷. Some clients already reduced spending by 20-30% in 2025⁷.

Where does this lead? The content creation that AI systems depend on (the journalism, the analysis, the tutorials, the documentation) loses its economic foundation. The golden goose is being processed into foie gras.

Some publishers are pursuing alternatives. Subscription models for those with loyal audiences. Licensing deals to become data suppliers rather than consumer-facing brands. Direct audience development that bypasses search entirely. But these options aren’t available to everyone, and they represent a fundamental restructuring of how information gets created and distributed.

Who Benefits, Who Suffers

The distribution of harm isn’t even. Premium publications with strong brand loyalty and subscription models (think The New York Times or The Wall Street Journal) may survive, even thrive in some ways. Their content is valuable enough that some users will pay directly.

But consider what gets lost: the mid-tier publication that does solid investigative work but lacks a household name. The niche blog that serves a small but dedicated community. The independent journalist building an audience through search visibility. The specialized tutorial site that answered your obscure programming question at 2 AM.

These are the long tail of information, the diversity and depth that made the web feel infinite. And they’re economically unsustainable in a zero-click world.

Meanwhile, the AI companies benefit enormously. They extracted training data from the open web, built products that replace web browsing, and capture the economic value that used to flow to publishers. It’s a classic platform play, except the platform is cannibalizing the ecosystem it depends on.

The Control Question

There’s another dimension to this that goes beyond economics.

When AI becomes the primary interface to information, the entities that control AI systems gain enormous influence over what people know and believe. This isn’t hypothetical. It’s the logical extension of the platform power we’ve already witnessed with social media.

Today, a handful of companies (OpenAI, Google, Anthropic, Meta) control the AI systems that millions rely on for information. Their choices about training data, fine-tuning, guardrails, and response generation shape what users receive. Even well-intentioned choices have consequences. Whose perspectives get amplified? Whose get suppressed? What counts as reliable information? These questions have always existed for search engines, but AI answers make them more consequential and less transparent.

And then there are the not-so-well-intentioned possibilities. State actors seeking to influence public opinion. Corporate interests shaping product recommendations. Political operatives nudging voters toward particular conclusions. The tools for information warfare become more powerful and more personal when they operate through trusted AI companions.

In my earlier piece on AI trust, I described how AI could influence thinking through what seems like personalized care, helping you “think through” issues while subtly emphasizing certain perspectives. The zero-click internet is the infrastructure that makes this possible at scale.

What Comes Next

I don’t have clean solutions. This is one of those situations where I find myself observing an emerging problem without obvious answers.

A few things seem clear:

The economics have to be restructured somehow. Whether that’s AI companies paying licensing fees to publishers, new forms of micropayments for content, or some other model, the current trajectory where AI extracts value without replenishment is unsustainable. The content has to come from somewhere, and its creation has to be funded. Related: Cloudflare’s Pay Per Crawl.

Transparency matters more than ever. When AI systems make curatorial choices about information, users deserve to understand how those choices are made. What sources are being prioritized? What biases exist in training data? What objectives are being optimized for? The “just trust us” model that worked (barely) for search engines won’t cut it for systems that synthesize and present information as finished answers.

Information diversity needs protection. The long tail of publishers, the niche sites, the independent voices, the specialized resources, represents something valuable about human knowledge. If zero-click economics drives them out of existence, we lose not just individual sources but the diversity of perspective that makes information environments healthy.

Users need to develop new habits. There’s value in occasionally clicking through to sources, even when the AI answer seems sufficient. It’s a way of keeping the information ecosystem alive, of checking AI interpretations against primary sources, of maintaining the habit of verification. Easy? No. But important.

The Hollowing

What strikes me most about the zero-click internet is how quietly it’s happening. There’s no dramatic moment, no obvious crisis. Just a gradual shift in behavior: people asking AI instead of searching, getting answers instead of visiting, consuming information without encountering its sources.

The websites are still there. The content still exists. But the connection between creators and consumers is thinning, and the economic flows that sustained creation are being redirected.

I keep coming back to the metaphor of hollowing. From the outside, the internet looks the same. Same websites, same search engines, same vast repository of human knowledge. But inside, something is being extracted. The vitality that came from traffic, engagement, and economic exchange is draining away.

We’re not losing the internet. We’re losing the reason anyone would create for it.

And by the time we fully understand what’s happened, I worry we’ll find ourselves in an information environment shaped entirely by AI systems that trained on a web that no longer sustains itself, inheriting the biases, gaps, and limitations of a snapshot in time, while the live ecosystem that could have corrected those flaws withers away.

The zero-click internet isn’t just about whether you click a link. It’s about whether the infrastructure of human knowledge survives the transition to AI-mediated information access. That’s the question we should be asking, before the answer becomes academic.

---

I’ve published The Internet Is Being Rewritten for Machines, Thanks to Cloudflare which is a good continuation to this article, following the thread on zero-click internet, focussing more on AI.

---

This article was written by me, a human. I used an LLM-powered grammar checker for final review.

Sources