Standfirst
For three decades, the internet rewarded those who created more information. Generative AI is overturning that bargain. As content becomes abundant and inexpensive to produce, economic power is shifting toward the companies that decide what billions of people see, trust and act upon. The defining competition of the AI era is no longer over producing information. It is over controlling how information reaches human attention.
Act I: The Signal
When Google began replacing lists of links with AI-generated answers, most users saw a better search engine.
Publishers saw the first large-scale break in the economic bargain that built the modern web.
For more than two decades, publishers, businesses and independent creators invested time and money producing information. Search engines organized that information and sent readers to it. The resulting attention generated advertising revenue, subscriptions, software sales and online commerce. Google became one of the world’s most valuable companies without having to create most of the information that made Search indispensable.
AI Overviews quietly changed that relationship.
Instead of directing users toward the web, Google increasingly began answering questions itself. The information still originated elsewhere. The user’s attention increasingly did not.
A 2025 Pew Research Center study found that users who encountered Google’s AI Overviews clicked a traditional search result in 8% of searches, compared with 15% when no AI Overview appeared. The precise percentages matter less than the direction of travel. When answers remain inside the platform, attention does too.
That distinction may prove more consequential than the feature itself.
For most of the internet era, producing information and capturing economic value remained closely linked. Better journalism attracted readers. Better tutorials generated search traffic. Better reviews produced affiliate sales. Discovery rewarded creation.
Generative AI weakens that relationship.
For the first time, information can be generated, summarized and recombined at extraordinary scale before users ever encounter the original source. As production becomes cheaper, discovery becomes more valuable. Economic power begins moving away from those who create information toward those who determine how it reaches people.
That is not simply another chapter in the history of search.
It represents a structural shift in the economics of the internet.
Act II: The Bargain That Built the Internet
Every communications revolution solves one scarcity while creating another.
The printing press reduced the scarcity of books but increased the importance of publishers capable of deciding what deserved to be printed.
Broadcast television multiplied programming while concentrating power in the handful of networks that controlled distribution.
The internet appeared to dismantle those bottlenecks. Anyone could publish. Distribution became almost free. Search engines connected billions of people with information that previously would have remained invisible.
Yet the internet never eliminated gatekeepers.
It simply replaced physical ones with digital ones.
Search engines, recommendation systems and social feeds became the infrastructure through which information reached audiences. Their role was powerful but fundamentally different from traditional publishers. They directed attention rather than producing the information that attracted it.
That distinction created one of the most productive commercial ecosystems in modern history.
Creators competed to produce useful information.
Search engines competed to organize it.
Users benefited from both.
Entire industries grew around this arrangement. Digital publishers invested in evergreen reporting because search traffic could compound for years. Software companies built libraries of documentation knowing customers would eventually discover them. Independent creators established viable businesses by answering questions too specialized for large media organizations to cover.
The incentives were not perfectly aligned, but they reinforced one another.
The more useful information existed online, the more valuable search became.
The more valuable search became, the greater the incentive to create useful information.
It was a virtuous cycle.
Generative AI changes the economics because it changes the destination, not merely the journey.
Act III: The Break
The idea that attention is scarce is not new.
In his 1971 essay Designing Organizations for an Information-Rich World, the economist and future Nobel laureate Herbert Simon made an observation that has become one of the defining insights of the digital age: “A wealth of information creates a poverty of attention.”
For half a century, that sentence described the demand side of the information economy.
Generative AI changes the supply side.
Until recently, information production remained tied to human effort. Every investigative report required reporters. Every advertising campaign required designers and copywriters. Every software manual, product description and customer support article required someone to create it.
Technology made production cheaper.
It did not fundamentally separate content from labour.
Large language models begin to do exactly that.
A single organization can now generate millions of words, thousands of product descriptions, personalized marketing campaigns, software documentation and customer responses at costs that would have been commercially unrealistic only a few years ago. Video, image and voice generation are rapidly moving in the same direction.
The economic significance lies less in AI’s creativity than in its productivity.
For the first time since the printing press, the supply of content is becoming increasingly independent of the supply of human creators.
That changes competition itself.
For most of the internet era, publishing more created more opportunities to be discovered. Newspapers expanded coverage. Retailers expanded catalogues. Companies invested heavily in content because each additional page increased the likelihood of attracting another visitor.
AI begins to erode that logic.
When every competitor can generate competent content at scale, producing one more article or one more marketing campaign becomes less valuable as a source of competitive advantage.
Human attention becomes the last scarcity, not because it is new, but because every technological advance that expands the supply of information makes attention relatively more constrained.
Something else therefore becomes scarce.
Not information.
Not production.
Discovery.
The internet democratized publishing.
Generative AI may re-centralize discovery.
Every major technological revolution changes what becomes abundant.
The steam engine made mechanical power abundant.
Electricity made energy abundant.
The internet made publishing abundant.
Generative AI is making content itself abundant.
Whenever abundance arrives, economic value does not disappear.
It migrates.
Creation scales with effort.
Allocation compounds.
Every search query, recommendation and AI conversation improves the next one, attracting more users and generating more behavioural data in a self-reinforcing cycle. Once content becomes abundant, allocation becomes a stronger economic moat than creation.
In the AI economy, value is beginning to migrate from those who create information toward those who determine which information reaches people.
Once that shift becomes visible, much of today’s AI race suddenly makes more sense.
Act IV: Where Economic Power Is Moving
This helps explain why the companies investing most aggressively in artificial intelligence share an important characteristic.
They are not simply producing more content.
They are positioning themselves to control how content is discovered.
Google’s investment in AI is not fundamentally about writing better answers. It is about ensuring that Search remains the primary gateway to information even if users stop clicking links. Alphabet has committed unprecedented levels of capital expenditure to AI infrastructure, custom silicon and data centres while embedding generative AI across Search, Workspace and Cloud. Investors are not financing a better chatbot. They are financing Google’s ability to remain the world’s default allocation engine.
The commercial logic is already becoming visible.
The Pew Research Center found that when Google’s AI Overview appeared, users clicked traditional search links at roughly half the rate of users who received conventional search results. Independent analyses have also documented the continued rise of zero-click searches, where users receive answers without ever visiting the websites that produced the underlying information.
For publishers, this is more than a traffic problem.
It is a redistribution of attention.
TikTok arrived at the same destination by a different route.
Earlier social networks primarily connected users with people they already knew. TikTok built its business around recommending content from people users had never met and often did not follow. Discovery no longer depended on relationships. Relationships increasingly emerged from discovery.
The recommendation engine became the product.
Generative AI amplifies that economic model.
When creating content becomes inexpensive, deciding which content deserves attention becomes more valuable.
OpenAI represents the next stage of that transition.
ChatGPT increasingly functions as an interface to information rather than another website competing for traffic. Users ask questions instead of opening multiple tabs. They request summaries instead of comparing competing sources. They increasingly begin their search for information inside an AI system rather than a search engine.
Every answer reinforces a subtle but important shift.
The economic advantage no longer lies solely in creating knowledge.
It lies in becoming the place where knowledge is encountered.
That helps explain where capital is flowing.
Alphabet continues expanding AI infrastructure at a scale measured in tens of billions of dollars annually.
Meta has repeatedly increased its capital expenditure guidance to accelerate AI infrastructure and recommendation systems.
Microsoft continues investing heavily in AI-enabled cloud infrastructure while embedding generative AI throughout Microsoft 365 and Azure.
These companies are not investing because AI allows them to generate more content.
The internet already democratized publishing.
AI industrializes it.
They are investing because AI is becoming the layer through which users discover, evaluate and act upon information.
That is where future economic power is expected to accumulate.
As producing information becomes increasingly commoditized, allocating human attention becomes increasingly strategic.
This is the structural transition hiding beneath the excitement surrounding generative AI.
The race is no longer simply to create better content.
It is to become the indispensable interface through which everyone else’s content is discovered.
Act V: The Implications
That shift carries implications far beyond publishers, search engines or AI companies.
It changes how competition itself should be understood.
For much of the past decade, regulators measured digital market power through familiar indicators: market share, pricing, defaults, app stores and platform access. Those concerns remain important.
AI introduces another source of power.
Control over discovery.
If AI assistants increasingly become the primary interface through which people discover information, products and services, then the competitive bottleneck is no longer search alone.
It is discovery itself.
The regulatory gap is becoming increasingly clear. Competition policy has become adept at examining who controls markets. It is only beginning to grapple with who controls the pathways through which markets are discovered.
That distinction matters.
Control over discovery influences which businesses are encountered, which publishers are read, which products are purchased and which software becomes the default choice. The companies allocating attention do not merely participate in markets.
Increasingly, they shape how markets are experienced.
The implications extend well beyond technology.
Publishers may produce exceptional journalism yet lose visibility if AI summaries satisfy readers before they visit the original reporting.
Retailers may discover that product quality matters less than whether recommendation systems surface their products.
Software companies will compete not only on features, but on whether AI assistants recommend their tools inside broader digital workflows.
Education faces the same transition.
Knowledge becomes easier to generate.
Judgment, trust and curation become harder to automate.
None of this diminishes the importance of original creation. It changes where value accumulates after creation occurs.
Throughout history, technological revolutions have rarely destroyed economic value.
They have relocated it.
The steam engine reduced the cost of movement.
Electricity reduced the cost of power.
The internet reduced the cost of publishing.
Generative AI is reducing the cost of creation itself.
When creation becomes abundant, value does not disappear.
It moves.
The defining companies of the next decade may not be those that create the world’s information.
They may be those that determine which information the world ever sees.
