The Macro Shift · established evidence

The Zero-Click Web: How AI Answers Ended the Referral Economy

Last reviewed 2026-08-11. Written by Chandranshu Kumar, Founder, Raveneye Global. · 11 min read

Every dominant medium has set the terms of the economy and the power built on top of it, and the open web is now living through its own version of that rule. For two decades, discovery ran through a search index: a business or a publisher wrote a page, ranked for it, and collected a click that carried an ad impression and, often, a sale. That click funded reporting, review sites, and millions of small publishers who had no other route to an audience. AI Overviews and chatbot answers have started to remove the click from the middle of that transaction. When Google shows an AI-written answer above the results, the click-through rate on the top organic page falls by nearly 60 percent, and global referral traffic to publishers is down by roughly a third in a year. The businesses answering the query still write the content the systems draw on. They are no longer reliably visited for it. Deciding who gets named now sits inside a small number of platforms, and that is a shift in economic and informational power, not just a traffic metric.

The referral economy the open web ran on

The printing press needed capital and a press. Broadcast needed a transmitter and a license. The open web needed neither: anyone could put up a page. What stayed scarce was not the ability to publish but the ability to be found, and for most of the last twenty five years that scarcity was resolved by a single layer sitting between a writer and a reader: the search index. A page that ranked for a searcher's query received a visit. A visit could carry an ad, a subscription pitch, or a sale. The arrangement was simple enough that an entire profession, and a large share of digital publishing revenue, grew up around earning that click.

It was also, from the start, a form of gatekeeping. Ranking well in one company's index became the precondition for being read at all, and that company set the rules, changed them without notice, and took a cut of the advertising built on top of the traffic it sent. But the exchange still had a floor: if a page ranked, the visitor arrived on that page, and the site captured the impression, the data, and the chance to convert them. Two decades of independent publishing, from national newsrooms down to single-author blogs, were built on that floor holding.

What has changed since 2023 is not the index. It is what sits above it. Google, OpenAI, and Microsoft now generate a synthesized answer and place it in front of the reader before any link appears, and a growing share of queries never reach a page at all. The floor that the referral-click economy stood on is being removed from underneath it, and the data on how fast that is happening is now large enough, and consistent enough across independent measurements, to read as a structural shift rather than a seasonal dip.

The click that stopped happening

The clearest single number comes from Ahrefs, which examined 300,000 keywords and found that when an AI Overview appears above Google's organic results, the click-through rate on the top-ranking page falls by nearly 60 percent. The searcher gets an answer synthesized from several sources, reads it on the results page, and has no remaining reason to open the page that supplied the facts. The visit that used to follow a search now often does not happen.

The effect shows up at scale across two independent measurements of publisher traffic. Global Google referral traffic to publishers fell by approximately 33 percent in the year to November 2025. Chartbeat, which tracks more than 2,500 news sites directly, recorded a 34 percent fall in Google search referrals between December 2024 and December 2025. Two separate methods, one from aggregate referral data and one from a real-time panel of newsrooms, arrive at the same order of magnitude within the same twelve months, which is what turns a single alarming figure into a corroborated trend.

Chatbots widen the gap further. By October 2025, ChatGPT had reached roughly 800 million weekly active users, up from about 300 million in December 2024. That is not search traffic redirected within Google's own results; it is a separate, fast-growing surface on which a question is asked and answered without a search engine, a results page, or a ranked link in the loop at all. Zero-click search, the practice of resolving a query without a click to any external site, is no longer a Google-specific phenomenon. It is becoming the default mode of an entire category of information-seeking.

Who is absorbing the loss

The decline is not spread evenly, and the pattern says something about who the referral economy actually protected. Digital Content Next found that small publishers, sites with 1,000 to 10,000 daily page views, lost 60 percent of their Google referral traffic over two years, against a 47 percent loss for mid-size publishers in the 10,000 to 100,000 range. The smaller a site, the more completely its distribution depended on a single channel it did not control, and the harder that channel's contraction lands. A large publisher with an app, a newsletter list, and a loyal direct audience has other doors. A small one, in most cases, had only the one.

The named casualties are large enough to be visible individually. Business Insider's organic search traffic fell 55 percent between April 2022 and April 2025, a decline that contributed to a 21 percent reduction in staff. HuffPost's desktop and mobile sites lost roughly half of their search referral traffic over the same shift toward AI-mediated search. BuzzFeed laid off approximately 180 employees, about a third of its workforce, in 2025, as the company reported losses of roughly $58 million a year and its stock fell from around $4 to under $0.50 a share.

None of these companies collapsed for one reason alone; advertising markets, strategy decisions, and years of accumulated pressure all played a part, and that caveat matters when reading any one company's numbers. What is harder to explain away is the consistency of the underlying traffic figures across publishers of very different size, ownership, and editorial focus, all falling in the same two-year window, tracked by different measurement firms using different methods. The referral-click model that funded these newsrooms is being drained faster than most of them can replace it.

Crawled without being paid

The same content that no longer earns a click is being read more than ever, just by machines instead of people. Cloudflare measured the ratio of pages an AI crawler scrapes against the visitors it sends back: roughly 14 to 1 for Google in June 2025, about 1,700 to 1 for OpenAI, and approximately 73,000 to 1 for Anthropic. Anthropic's crawlers, in other words, read tens of thousands of pages for every reader they returned to the sites that wrote them. Google, still tied to sending some traffic to preserve its own results page, remains the most balanced of the three by a wide margin.

The imbalance is accelerating on the crawling side even as it worsens on the referral side. GPTBot's raw crawl requests grew 305 percent year on year by May 2025, and overall AI crawler traffic across the web rose 65 percent in six months. The two curves are moving in opposite directions at the same time: more of a site's content is being ingested to train models and answer questions, and less of its former audience is being sent back to see it there.

Publishers have started to respond at the infrastructure level rather than waiting on individual deals. Cloudflare opened its Pay Per Crawl system, which lets a site charge an AI crawler for access instead of blocking it outright, to all paid customers on August 28, 2025. By late 2025, customers on the service were already sending more than 1 billion HTTP 402 "payment required" responses to crawlers every day. It is an early and largely one-vendor mechanism, not yet a settled market price for a crawl, but it is the first attempt to put a toll booth where an open gate used to be.

Who controls the answer

For most of the search era, the power sitting between a writer and a reader was intermediation: a company decided who ranked, but the reader still had to leave that company's page to reach the source. That single step, however imperfect, meant the site being written about kept a relationship with its own audience. What has changed is that the intermediary now also generates the final answer and delivers it end to end, without the reader needing to leave at all. The number of companies capable of doing that at scale, Google, OpenAI, Microsoft, and a handful of others, is smaller than the number of sites that ever ranked in a search index, and it is a narrower, more concentrated set of hands than the one the referral economy grew up under.

That concentration has a geography. The platforms assembling and delivering the answer are overwhelmingly headquartered in the United States, while the content they draw on to write that answer is produced everywhere: newsrooms in Mumbai and Manila, trade publications in London, product review sites run out of a spare room in Ohio. The question of who gets compensated, and who gets cited by name, for information written anywhere in the world is increasingly being decided inside a small number of corporate systems built by companies based in one country. That is not a new pattern in the history of dominant media, but it is a new and rapidly hardening instance of it, and it sits underneath what looks on the surface like a purely technical shift in how search results are displayed.

Two things hold at once. The referral-click web itself was never a level playing field; a single company's ranking algorithm already decided who could be found, and it took a share of the advertising built on top of the traffic it sent. Answer engines have not invented gatekeeping over discovery. They have moved that gatekeeping one step further downstream, from deciding which page a reader sees to deciding what a reader is told, and removed the one part of the old bargain, the click, that gave the source any direct return.

What replaces the click

Three responses are underway at once, and none of them has settled into a standard yet. Pay Per Crawl and systems like it try to price access to content at the point of the crawl rather than the point of the click, a toll rather than a referral. Direct licensing deals between publishers and AI companies are being negotiated case by case, outside public view in most instances, and the terms available to a large newsroom with legal leverage look nothing like the terms, if any, available to a small independent site. Neither approach yet reaches most of the small and mid-size publishers who, by Digital Content Next's figures, are losing traffic fastest.

For a business that cannot negotiate a licensing deal or does not run its own crawl-metering infrastructure, the practical adaptation is narrower: be legible enough, in structure and in the facts a page states plainly, that when an engine does draw on it for an answer, that business is the one named in the citation rather than folded anonymously into a synthesized paragraph. A citation is not a click, and it does not restore the old revenue, but it is the only form of credit available on a surface that was not built to send visitors anywhere.

Whether this settles into a licensed, compensated relationship between the platforms generating answers and the sites supplying the facts inside them, or whether the open web's smaller and mid-size layer keeps shrinking under a system that owes it neither a click nor a credit, is not yet decided by the numbers in hand. What the numbers already show, measured independently by at least four separate organizations across two years, is that the click-funded model built the last two decades of digital publishing is contracting in real time, and nothing in the current data suggests it reverses on its own.

The evidence

Key findings, with their sources

When an AI Overview appears above Google's organic results, the click-through rate on the top-ranking page falls by nearly 60 percent, in an analysis of 300,000 keywords.

established Ahrefs data cited in Digiday (2025).
Global Google search referral traffic to publishers fell approximately 33 percent in the year to November 2025.

established Press Gazette, media audience and business data trends report (2025/2026).
Google Search referrals across more than 2,500 news sites tracked directly fell 34 percent between December 2024 and December 2025.

established Chartbeat data, reported by Press Gazette (2026).
Small publishers (1,000 to 10,000 daily page views) lost 60 percent of Google referral traffic over two years, against a 47 percent loss for mid-size publishers (10,000 to 100,000 daily page views).

established Digital Content Next (2025).
Business Insider's organic search traffic fell 55 percent between April 2022 and April 2025, contributing to a 21 percent staff reduction; HuffPost's desktop and mobile sites lost roughly half of their search referral traffic over the same shift.

established AdExchanger (2025).
BuzzFeed laid off approximately 180 employees, about a third of its workforce, in 2025, reporting annual losses of roughly $58 million as its stock fell from about $4 to under $0.50 a share.

established Futurism, citing Hollywood Reporter data (2024-2025).
Cloudflare measured crawl-to-referral ratios of roughly 14 to 1 for Google, 1,700 to 1 for OpenAI, and 73,000 to 1 for Anthropic, meaning Anthropic's crawlers scraped tens of thousands of pages for every one visitor sent back.

established Cloudflare Radar data (June 2025).
ChatGPT reached 800 million weekly active users by October 2025, up from 300 million in December 2024, a scale of direct-answer traffic diverted from the open web in under a year.

established OpenAI usage disclosures, via industry trackers (2025-2026).

Calibration

What is proven, what is promising, what is unproven

Evidence tier	Tactics	What the evidence says
established	The core measurements: the AI Overview click-through collapse, the year-on-year Google referral declines across large publisher panels, and the crawl-to-referral imbalance recorded by Cloudflare.	Corroborated across at least four independent organizations (Ahrefs, Press Gazette, Chartbeat, Digital Content Next, Cloudflare) using different methods, over the same two-year window, all pointing the same direction.
emerging	Attributing specific company outcomes, layoffs, stock declines, and shutdowns, directly and solely to AI-search traffic loss, and the durability of crawl-metering tools like Pay Per Crawl as a lasting compensation mechanism.	Individual cases (BuzzFeed, Business Insider, HuffPost) are well documented, but each sits alongside confounding factors such as ad-market conditions and strategy pivots; Pay Per Crawl adoption data is early, drawn from one vendor, and not yet independently audited.
contested	Whether the referral collapse is a temporary transition cost of a platform shift that resolves through licensing and new distribution, or a permanent restructuring of how the open web is funded.	No public dataset yet tracks whether affected publishers recover through licensing deals, direct and app-based distribution, or a different monetization model entirely; informed parties argue both continuation and reversal, and the period measured so far is too short to settle it.

Reference

Glossary

Zero-click search: A search that is resolved on the results page or inside a chatbot answer itself, with no click to any external site required to satisfy the query.
Referral traffic: Visits to a website that arrive from a link on another site or a search engine's results page, as distinct from direct visits or app traffic.
Click-through rate (CTR): The share of people who see a result, whether a search listing or an AI Overview citation, and go on to click through to the underlying page.
Crawl-to-referral ratio: The number of times an automated crawler reads a site's pages for every one visitor that same platform sends back to the site, used to measure how much a platform gives back relative to what it takes.
Pay Per Crawl: An infrastructure-level system, launched by Cloudflare in 2025, that lets a website charge an AI crawler for access to its pages instead of blocking the crawler outright.
AI Overview: A synthesized, AI-written answer that Google places above the standard list of organic results, generated from multiple source pages and displayed before any link.

Straight answers

Frequently asked questions

What is the zero-click web?

It is a search environment in which a growing share of queries are answered directly on the results page or inside a chatbot, with no click to any outside website required. Ahrefs found that when Google shows an AI Overview, the click-through rate on the top organic result falls by nearly 60 percent, meaning the searcher's question is often resolved before any site is visited.

How much has AI-mediated search actually cut publisher traffic?

Global Google referral traffic to publishers fell approximately 33 percent in the year to November 2025, and Chartbeat's panel of more than 2,500 news sites recorded a 34 percent fall over roughly the same period. Small publishers were hit hardest, losing 60 percent of referral traffic over two years against 47 percent for mid-size sites.

Which publishers have been hit hardest?

Smaller publishers with fewer alternate distribution channels absorbed the largest losses. Named cases include Business Insider, whose organic search traffic fell 55 percent between April 2022 and April 2025 alongside a 21 percent staff reduction, HuffPost, which lost roughly half its search referral traffic, and BuzzFeed, which laid off about a third of its workforce in 2025.

Are AI companies paying publishers for the content they use?

Not as a general practice yet. Cloudflare's crawl data shows AI companies reading far more pages than they send visitors back to, roughly 1,700 pages per visitor for OpenAI and about 73,000 for Anthropic. Some publishers now charge crawlers directly through systems like Cloudflare's Pay Per Crawl, and some larger publishers have negotiated private licensing deals, but neither reaches most small and mid-size sites yet.

Will referral traffic recover?

That is not settled by the current data. The decline is measured consistently across independent sources over two years, but whether it stabilizes through licensing and new compensation mechanisms, or continues as the answer layer keeps growing, is an open question that the numbers so far do not resolve either way.

Provenance

Sources

Digiday, "Google AI Overviews linked to 25% drop in publisher referral traffic, new data shows" (Ahrefs analysis of 300,000 keywords) (2025)digiday.com
Press Gazette, "Google traffic down: 2025 trends report" (2025/2026)pressgazette.co.uk
Chartbeat data on more than 2,500 news sites, via Press Gazette (2026)pressgazette.co.uk
Digital Content Next, "Facts: Google's push to AI hurts publisher traffic" (2025)digitalcontentnext.org
AdExchanger, "The AI Search Reckoning Is Dismantling Open Web Traffic, and Publishers May Never Recover" (2025)adexchanger.com
Futurism, citing Hollywood Reporter data on BuzzFeed layoffs and financial results (2024-2025)futurism.com
Cloudflare Radar, "From Googlebot to GPTBot: who's crawling your site in 2025" (crawl-to-referral ratios, GPTBot growth) (2025)blog.cloudflare.com
Cloudflare Blog, Pay Per Crawl launch and adoption data (2025)blog.cloudflare.com
OpenAI usage disclosures on ChatGPT weekly active users, via industry trackers (2025-2026)

Every figure above is attributed to a real, dated source and tagged with its evidence tier. Where a claim could not be verified to a primary source, it is not stated as fact.