Introduction: From Search Engine to Answer Machine
For most of the internet’s history, search engines occupied a relatively humble role. They acted as guides rather than destinations. Users typed a question, received a list of links and chose which source to trust. Publishers, businesses and independent creators built websites on the assumption that discovery would eventually lead readers to their door. The arrangement became one of the defining economic bargains of the digital age. Google organized the world’s information. Publishers created it. Advertisers paid for access to audiences. Everyone complained about the system, but most participants benefited from it.
Today, that relationship is changing. Artificial intelligence has made producing content easier than at any point in history. Articles, product descriptions, search engine optimization and advertising campaigns can be generated in seconds. Yet while the supply of content has exploded, the mechanisms that once distributed it are undergoing a profound transformation.
Search engines are increasingly becoming answer engines.
Rather than directing users toward websites, platforms are beginning to provide responses themselves. AI-generated summaries sit above traditional search results. Chatbots answer questions without requiring users to visit original sources. Consumers who once navigated from page to page now receive information through a single interface.
The consequences extend beyond Google. OpenAI, Microsoft, Perplexity and others are competing to become the primary gateway through which information is consumed. Their products differ, but they share a common characteristic. They seek to reduce friction. Users no longer need to sift through dozens of links. The answer arrives immediately. From the perspective of consumers, the experience is faster and often more convenient. From the perspective of publishers, the economics become less clear.
For nearly three decades, the web functioned as a giant marketplace of links. Traffic was the currency. Visibility could be converted into advertising revenue, subscriptions, affiliate commissions or sales. Search engine optimization became a multibillion-dollar industry because appearing on the first page of Google meant being discovered. But if users stop clicking, the value of being found changes.
This raises a more fundamental question. If AI systems increasingly act as intermediaries between people and information, what happens to the businesses and institutions that produce the information in the first place? The answer may determine whether the next era of the internet remains open, or evolves into something more centralized and less dependent on the link itself.
The Vanishing Click
The decline of web traffic is not theoretical. It is already visible in the data. According to research by SparkToro founder Rand Fishkin, approximately 58.5 per cent of Google searches in the United States end without a click to an external website. In Europe, the figure approaches 60 per cent. Out of every 1,000 searches performed, only around 360 result in visits to the wider web.
The rest remain inside Google’s ecosystem. Some searches lead users to YouTube videos, Google Maps listings or Google-owned properties. Increasingly, many queries are answered directly on the search page itself. Sports scores, weather forecasts, definitions and restaurant information no longer require a visit elsewhere. The trend has become known as the zero-click search.
For years, zero-click results were concentrated around simple questions. Looking up the capital of France or checking tomorrow’s weather hardly required sending users elsewhere. But artificial intelligence has expanded the scope of queries that can be answered immediately. The implications become more significant when complex questions are involved.
A user researching financial concepts, travel recommendations or consumer products may now receive a synthesized response drawn from several websites without ever opening them. The information is extracted from publishers, condensed and presented directly within the platform. The shift appears modest at first glance. After all, consumers still receive answers. But the economics of publishing depend not on answers, but on attention.
HouseFresh, a British website specializing in air purifier reviews, illustrates the challenge. Founded by entrepreneur Danny Ashton, the site built its reputation through detailed testing and independent evaluations. For years, it ranked prominently in Google search results. Following algorithm changes in 2023 and 2024, traffic declined by more than 90 per cent. HouseFresh was hardly alone.
Across thousands of niche websites, similar stories emerged. Hobby blogs, specialist review sites and independent publishers found themselves displaced by larger platforms, user-generated forums and Google’s own increasingly comprehensive search features. The problem was not necessarily that readers migrated elsewhere. Often, they simply stopped clicking.
That distinction matters. Throughout the history of the web, traffic moved between winners and losers. One site gained prominence while another declined. Yet traffic itself remained abundant. Today, the total pool appears to be shrinking.
The internet has not run out of information. On the contrary, it has never contained more. What appears to be changing is the role of websites themselves. Increasingly, content is becoming raw material rather than destination. For the first time since the commercial internet emerged in the 1990s, producing information and capturing attention are becoming separate businesses.
Google’s Code Red
For more than two decades, Google enjoyed a position few companies in history have attained. Its name became a verb. To search for information was, quite literally, to “Google it.” That dominance was built on a deceptively simple model. Users arrived with questions. Google directed them to websites. Advertisers paid for the privilege of appearing alongside those searches. In 2024, Google’s parent company Alphabet generated more than $350 billion in revenue, with advertising accounting for roughly three quarters of the total.
The arrangement proved remarkably durable. Search advertising became one of the most profitable businesses ever created. Then came ChatGPT. When OpenAI released its chatbot in November 2022, few expected the speed with which it would capture public attention. Within two months, ChatGPT had reached 100 million monthly users, making it one of the fastest-growing consumer applications in history. Inside Google, the response was immediate.
According to reporting by The New York Times, executives declared an internal “Code Red.” Founders Larry Page and Sergey Brin were brought back into discussions. Engineers were reassigned. Projects previously considered experimental suddenly became urgent. The concern was not that ChatGPT had surpassed Google’s search quality. Rather, it threatened the assumptions underpinning Google’s business. Traditional search follows a sequence that has remained largely unchanged since the late 1990s. Users search, examine results and click through to external sites. Advertisements are displayed along the way.
Chatbots operate differently. Users ask a question and receive an answer. There are no lists of blue links. There are fewer opportunities to display advertisements. Most importantly, there is no guarantee that users ever leave the interface. For Google, this represented what Harvard professor Clayton Christensen famously described as the innovator’s dilemma. The company faced competition from a new technology that offered a different experience, even if it initially appeared less profitable. Ignoring the threat carried obvious risks. Embracing it threatened Google’s own business model.
The irony was difficult to miss. Google had spent years perfecting a system that rewarded sending users elsewhere. Suddenly, the future of search appeared to involve keeping users inside. By 2024, Google had integrated Gemini across much of its product ecosystem. AI-generated summaries began appearing above conventional search results. The company later introduced AI Mode, which transformed search into something closer to a conversational assistant.
Google was hardly alone. Microsoft integrated OpenAI technology into Bing and Copilot. Perplexity AI positioned itself as an answer engine rather than a search engine. Anthropic and Meta entered the race with their own models. What emerged was less a competition over websites and more a competition over interfaces.
For three decades, browsers and search engines served as gateways to information. Increasingly, the gateway itself is becoming the destination. Consumers appear comfortable with the change. ChatGPT surpassed 800 million weekly active users in 2025, while Google’s AI Overviews now reach more than two billion users globally. Convenience tends to win.
Yet convenience can alter incentives in unexpected ways. The web grew because millions of individuals and organizations had reasons to publish information. Traffic translated into economic value. Search engines benefited because they organized that information. Artificial intelligence introduces a different relationship. Instead of acting primarily as librarians, platforms are beginning to function more like editors, synthesizing knowledge from many sources and presenting a single answer. For users, this may represent progress. For publishers, the question becomes whether the economics of creating information remain intact when distribution itself is increasingly concentrated.
When Answers Stay at Home
Google’s introduction of AI Overviews in 2024 marked one of the most significant changes to search since the company’s founding. The concept was straightforward. Rather than displaying a list of websites, Google would summarize information directly on the results page, generating responses synthesized from multiple sources. From a user’s perspective, the appeal was obvious. The answer appeared instantly. There was no need to compare ten different articles or navigate through pages crowded with advertisements and pop-ups.
For publishers, however, the implications were less encouraging. Research published by the Pew Research Center in 2025 tracked nearly 69,000 Google searches conducted by 900 American adults. The findings illustrated how dramatically user behavior changes when AI summaries appear. When an AI overview was present, users clicked on external websites only 8 per cent of the time. Without the summary, the figure rose to 15 per cent. Even links embedded within AI Overviews attracted remarkably little engagement, with only around 1 per cent of users clicking through. In other words, people consumed information while bypassing the places that produced it. The effect resembles window shopping without entering the store.
Publishers still provide the underlying material. Their articles, reviews and analysis become ingredients used to generate responses. Yet the customer increasingly leaves with the answer before reaching the source. For decades, Google’s interests aligned with those of content creators. Better websites meant better search results. Better search results attracted more users. More users generated more advertising revenue.
Artificial intelligence complicates that relationship. The answer itself becomes the product. Google insists that AI Overviews drive higher-quality traffic and encourage more sophisticated queries. The company argues that users engage with content differently rather than abandoning it altogether. Publishers remain unconvinced. Media executives have compared the experience to watching traffic disappear without understanding precisely where it has gone. The concern is not simply declining page views. It is the possibility that the underlying economics of discovery are being rewritten.
The impact is uneven. Large platforms with strong brands continue to attract audiences directly. The New York Times, YouTube and Wikipedia benefit from recognition and scale. Smaller publishers face a more uncertain future. For years, expertise alone could earn visibility. A niche website devoted to camera lenses, air purifiers or cycling equipment could compete through quality and patience. Artificial intelligence changes the equation. When answers remain inside the platform, being the best source matters less than being part of the training data. And for many creators, being an ingredient is considerably less valuable than being a destination.
The New Winners
Every technological shift produces beneficiaries as well as casualties. The rise of AI search is no exception. Contrary to fears that artificial intelligence would make the web irrelevant overnight, traffic has not disappeared evenly. Instead, it has become increasingly concentrated. Wikipedia remains among the biggest winners. So does YouTube, which Google owns. Reddit, once dismissed as an unruly collection of forums and anonymous discussions, has emerged as one of the internet’s most valuable reservoirs of human-generated content. Their advantage lies not necessarily in superior expertise, but in scale and familiarity.
Users searching for recommendations often append “Reddit” to their queries. Google itself increasingly prioritizes forum discussions, user experiences and conversational content. A decade ago, many of these same pages would have been regarded as low-quality results. Today, they are prized because they contain what artificial intelligence systems struggle to generate convincingly: genuine human interactions. The shift has been dramatic. According to SEO platform Sistrix, Reddit’s visibility in Google search increased by more than 1,300 per cent between mid-2023 and early 2024. The number of Reddit pages appearing prominently in search results nearly doubled from around 22 million to more than 41 million. Traffic followed.
By 2024, Reddit had become one of Google’s largest referral destinations in the United States, rivaled only by Wikipedia and YouTube. The relationship deepened in February 2024 when Google signed a licensing agreement reportedly worth about $60 million annually, granting access to Reddit’s data for AI training purposes. Shortly afterward, Reddit restricted access to many third-party crawlers, effectively ensuring that Google maintained privileged access to one of the internet’s richest collections of conversations. The arrangement reflected a broader shift taking place across the web. In the search era, authority was earned largely through links. Websites accumulated reputation over time. Smaller publishers could compete through specialization and expertise. The answer era rewards different characteristics. Scale matters. Familiarity matters. User-generated content matters.
Smaller sites often struggle to benefit from these trends. HouseFresh, the independent review website founded by Danny Ashton, became a case study in this new reality. Despite conducting extensive product testing and publishing original research, the site experienced traffic declines exceeding 90 per cent following algorithm changes. The irony is difficult to ignore. At a moment when artificial intelligence makes information abundant, original expertise appears to matter less in determining visibility.
Recognition increasingly outweighs specialization. This does not necessarily imply deliberate favoritism. Platforms optimize for engagement and trust signals that are easiest to measure. Large brands naturally possess more of both. The result, however, is a form of concentration. The internet once allowed niche publishers to prosper because search engines lowered barriers to discovery. A website devoted to espresso machines or hiking boots could compete with multinational companies if its information was sufficiently useful. Today, the advantages increasingly belong to institutions large enough to become destinations in their own right. The rest risk becoming invisible.
The Economics of Attention
The internet’s business model has always rested upon a deceptively simple principle. Attention pays.
A visitor arriving through Google might read an article, click an advertisement, subscribe to a newsletter or purchase a product. Individually, each interaction generated little revenue. Collectively, billions of visits sustained a vast ecosystem of publishers, retailers and independent creators. The scale was extraordinary. Global digital advertising spending exceeded $700 billion in 2025. Google and Meta alone accounted for more than half of that total. Thousands of websites, from local newspapers to specialist blogs, depended on search traffic to capture a small share of those advertising dollars. Artificial intelligence threatens to disrupt that equation.
When users receive answers without clicking, the economics change. Research company Raptive estimated that AI-generated summaries could eventually reduce publisher revenues by roughly $2 billion annually. Depending on the category, some websites could experience traffic declines of between 20 and 60 per cent. Those figures represent more than a financial inconvenience. Publishing carries fixed costs. Journalists, editors, product testers and researchers still require salaries. Investigations still take time. Expertise remains expensive. What disappears is not the cost of producing information, but the mechanism that once paid for it. Local newspapers provide an illustration.
Many had already struggled for years against declining print subscriptions and the migration of advertising to digital platforms. Search traffic became one of the few remaining sources of audience growth. If that traffic diminishes further, the economics become increasingly difficult. The same applies to independent websites. A product review site may spend weeks testing equipment. A travel blog may invest years building expertise. A medical publication may rely upon specialists to ensure accuracy. Yet artificial intelligence can summarize their conclusions in seconds. From the consumer’s perspective, this appears efficient. From the creator’s perspective, it resembles a peculiar form of disintermediation.
The value created by expertise is captured elsewhere. Historically, technology companies and publishers enjoyed a symbiotic relationship. Search engines benefited from high-quality information. Publishers benefited from traffic. Artificial intelligence introduces a subtle but profound shift. Content producers increasingly resemble suppliers rather than partners. The platforms own the interface. The publishers provide the raw material. This distinction matters because suppliers possess far less bargaining power. Large publishers such as The New York Times or Reuters can negotiate licensing agreements. Smaller creators often cannot. As a result, the economics of information may become increasingly concentrated among institutions large enough to command payment and platforms large enough to distribute it. Everyone else faces a more uncomfortable question.
If traffic no longer arrives, what exactly are they optimizing for?
The Paradox of AI
The technology industry frequently describes artificial intelligence as an accelerator of knowledge. In many respects, it is. Large language models allow people to summarize documents, answer questions and navigate information more efficiently than ever before. Tasks that once required twenty minutes can now be completed in twenty seconds. Yet the systems themselves depend upon a resource they do not create. Human knowledge.
Artificial intelligence models are trained on books, articles, discussions, videos and websites produced by millions of individuals over decades. The quality of their outputs depends upon the quality of their inputs. That creates a paradox. The more effective AI systems become at answering questions directly, the fewer incentives may exist for humans to continue producing the underlying material. Google, OpenAI, Microsoft and Anthropic all rely upon an ecosystem that predates them. Their products are only as useful as the information they can access. For now, the reservoir remains vast.
But reservoirs require replenishment. If independent publishers disappear, if specialist communities shrink and if original reporting becomes uneconomic, future AI systems may increasingly train on synthetic material generated by previous generations of AI. Researchers have already warned about the phenomenon known as model collapse, in which systems trained upon their own outputs gradually lose diversity and accuracy. Information ecosystems are remarkably resilient. They are also surprisingly fragile.
The internet flourished because millions of people had reasons to contribute. The question facing the answer era is whether those incentives survive when fewer people are rewarded for creating what everyone else consumes. For most of its history, the internet operated as an unusually open system.
A teenager with a blog could reach the same audience as a multinational corporation. A specialist forum dedicated to fountain pens or mountain bikes could appear above larger competitors if its information proved more useful. Hyperlinks connected millions of independent websites into something resembling a public commons. The web was chaotic and often inefficient, but discovery remained relatively decentralized. Artificial intelligence may be changing that. Increasingly, users interact with information through a handful of interfaces. Google, ChatGPT, Perplexity, Copilot and Meta AI are competing to become the starting point for knowledge itself. The destination matters less when the answer arrives before the journey begins.
In some ways, the trend resembles earlier shifts in media. Cable television reduced thousands of local channels into a handful of networks. Streaming services concentrated attention among a small number of platforms. Social media centralized conversations that once occurred across independent websites. The web may be undergoing a similar consolidation.
The danger is not necessarily censorship or deliberate manipulation. Rather, it is concentration. Discovery becomes mediated by algorithms whose priorities remain largely invisible. The pathways between creator and audience become narrower. Traffic accumulates around the largest players, while smaller voices struggle to attract attention. At the same time, the web itself is changing.
Low-cost AI tools have unleashed an explosion of synthetic content. Websites optimized entirely for search rankings can now generate hundreds of articles daily. Product reviews are written without products being tested. Travel guides are assembled by machines that have never visited the places they describe. Researchers and search engineers increasingly refer to this phenomenon as “AI slop.”
The problem is not simply that AI-generated content exists. Much of it is useful. The problem is that quantity overwhelms quality. For search engines and AI systems, distinguishing expertise from imitation becomes harder when both are produced at scale. Consumers have begun responding in curious ways. Many users now add “Reddit” to searches because they seek evidence of human experience. Others subscribe to newsletters or follow individual creators rather than relying on search. Communities that offer authenticity command a growing premium precisely because they are difficult to replicate.
The internet, paradoxically, may become more personal as it becomes more centralized. Yet the long-term risks extend beyond discovery. Artificial intelligence systems depend upon fresh information. News events, scientific breakthroughs and specialist expertise cannot be generated from existing data alone. They must originate somewhere. Researchers have warned about model collapse, the possibility that AI systems trained increasingly on synthetic outputs become less diverse and more prone to error. In a 2024 study published in Nature, researchers demonstrated that repeated training on AI-generated content can gradually degrade model quality.
The prospect remains theoretical, but the underlying logic is straightforward. A photocopy of a photocopy eventually loses detail. Information ecosystems face similar challenges. If independent reporting declines, if specialist communities disappear and if economic incentives weaken, future systems may find themselves consuming increasingly derivative material. The internet risks becoming less a library and more an echo chamber.
Paying for Knowledge
Publishers have begun searching for alternatives. The assumption that traffic alone can sustain digital media has weakened over the past decade. Artificial intelligence may accelerate that transition.
Some companies are experimenting with subscriptions. Others are pursuing licensing agreements with AI firms. The New York Times has filed lawsuits. Axel Springer has signed partnerships. Reddit and Shutterstock have negotiated agreements worth tens of millions of dollars. The logic increasingly resembles the entertainment industry. Music streaming services pay royalties. Television networks license content. Newspapers and publishers argue that AI companies should compensate creators whose work forms the foundation of their products.
Supporters view licensing as a necessary evolution. Critics warn that it could create a more fragmented and unequal information ecosystem. Large institutions possess leverage. Reuters, The Associated Press and major newspapers can negotiate. Smaller websites and independent writers often lack the scale necessary to secure meaningful agreements. The result may be a two-tier internet.
Premium publishers sell access to AI companies. Smaller creators receive neither traffic nor compensation. Meanwhile, audiences themselves are changing. Consumers increasingly trust personalities over institutions. Podcasts, newsletters and YouTube channels attract loyal followings that traditional media struggles to replicate. Direct relationships matter more when search traffic becomes unreliable.Some creators may thrive precisely because they cultivate communities rather than page views. In that sense, artificial intelligence could reinforce trends that were already underway. The internet may become smaller, more concentrated and more relationship-driven. But whether it remains economically sustainable is another question entirely.
Conclusion: The End of the Link Economy
The internet was built upon links. For three decades, hyperlinks served as both infrastructure and currency. They connected ideas, rewarded expertise and allowed unknown creators to find audiences. Search engines flourished because they organized this vast network rather than replacing it. Artificial intelligence represents a different model. Instead of directing users toward information, it increasingly delivers information directly. The experience is faster and often more convenient. Consumers benefit from reduced friction. Companies compete to provide ever more seamless answers.
Yet convenience can obscure trade-offs. Information does not appear spontaneously. It is researched, written, tested and refined by people and institutions whose incentives depend upon being discovered. If answers remain inside platforms, if clicks become scarcer and if traffic no longer rewards expertise, the economics underpinning the open web may gradually weaken.
This does not mean the internet is disappearing. Nor does it imply that Google, OpenAI or other companies are acting out of malice. Each is responding rationally to competitive pressures and consumer demand. But systems shaped by rational decisions can still produce unintended consequences. The original bargain of the web was remarkably simple. Publishers created content. Search engines guided readers. Advertisers paid the bills.
That bargain is being renegotiated.
The next era of the internet may prove more efficient, more conversational and more centralized. Users may spend less time searching and more time receiving answers. Businesses may rely less on websites and more on platforms. The question is whether the institutions that generate knowledge can continue to prosper in a world where distribution itself has become concentrated.The answer will determine whether artificial intelligence becomes the next chapter of the open web or the beginning of something fundamentally different. For decades, the internet rewarded those who built the best library. Increasingly, the race is to own the librarian.