The Future of the Web: The Web Becomes a Verb

For most of its history, the World Wide Web behaved like a place. We opened a browser, followed links, compared information, filled out forms, and completed whatever task had taken us online. Smartphones changed where that happened and broadband made it richer, but the relationship stayed familiar: the Internet supplied information and services while the user supplied the judgment, coordination, and effort needed to produce an outcome.

That division of labor is changing as artificial intelligence moves beyond answering questions and into getting things done. Software agents can search across services, interpret instructions, compare alternatives, interact with external systems, and carry portions of a task toward completion. Emerging protocols let agents use tools, communicate, transact, and assemble interfaces around particular objectives.

The Web is becoming a verb.

Websites will not disappear, nor does the transition require another label such as “Web 4.0.” Over the next five to ten years, users are likely to spend less time explaining how to accomplish something and more time stating what they want accomplished. The Web moves from information architecture toward execution architecture.

About 6 billion people, or 74 percent of humanity, were online in 2025, while 2.2 billion remained offline. The Internet is basic economic infrastructure. Its next transformation begins when that infrastructure interprets instructions rather than merely carrying them.

Sources: Pew Research Center, National Bureau of Economic Research

From Navigation to Intention

Planning a simple vacation reveals how much work users still perform. A traveler searches airlines and hotels, reconciles calendars, calculates a budget, and moves information between unrelated services. Economically, those steps create friction because the user becomes the integration mechanism between systems never designed to work together.

An intelligent Web begins with the objective instead. A traveler might ask for somewhere warm for four nights in February, preferably nonstop, for about $3,000. The system can determine what information matters and which services or constraints are relevant. In 2026, roughly 49 percent of U.S. adults used AI chatbots, 24 percent used them daily, and around four in ten used them to search for information. Among employed adults, 38 percent reported using them for work tasks. Conversational computing has entered ordinary behavior even if broad autonomous delegation has not.

The economics extend beyond interface. AI agents could reduce the friction involved in finding sellers, comparing prices, communicating, and contracting. Over time, those changes could also affect how markets are organized. Traditional UX asks how to make a process easier to navigate. Intelligent UX asks whether the user should navigate much of it at all.

Sources: International Telecommunication Union, Pew Research Center

The Interface Stops Being the Destination

Once intention becomes the starting point, the interface becomes less fixed. Traditional Web design assumes the path is built before the user arrives. AI can reverse that order by understanding the objective first and presenting only the information or controls required. Experimental systems such as Google’s A2UI point toward interfaces assembled dynamically around a task rather than forcing every task through permanent screens.

The screen will not vanish. A multi-interface Web is more plausible, with the smartphone central while intelligence spreads elsewhere. U.S. smartphone ownership reached 91 percent, up from 35 percent in 2011, while 35 percent of adults reported an AI-enabled smart speaker in 2026.

What looks increasingly wireless still depends on heavily wired infrastructure. Fixed broadband carried about 7.3 zettabytes of traffic in 2025 compared with 1.5 zettabytes over mobile broadband, and more than 99 percent of international data flows still moved through submarine cables. Ambient devices can become more personal and context-aware while the underlying network and computing infrastructure carries much of the computational burden.

From Behavior to Context

Greater intelligence requires richer context. The commercial Web learned from searches and purchases as well as broader patterns of online activity; connected environments extend observation into physical circumstances. In 2026, 35 percent of American adults reported an AI-enabled smart speaker, 18 percent a smart doorbell, 13 percent a robot vacuum, and 11 percent a smart thermostat.

Economic value increasingly comes from connecting those signals rather than merely accumulating them. An energy system becomes more useful when it combines an electric vehicle’s battery state with electricity prices and departure time. The first Web primarily knew what people did online. The next Web increasingly knows what is happening around them.

That usefulness carries a sharper bargain over control. Sixty-two percent of Americans believed AI analysis of personal information could make life easier, yet 70 percent had little or no trust in companies to use AI responsibly. Eighty percent expected information to be used beyond its original purpose, and 81 percent expected uses they would find uncomfortable. Richer intelligence does not technically require permanent centralized ownership of every signal; systems can instead limit what they receive through selective disclosure, local processing, or digital credentials.

AI also shifts the privacy problem from records toward inference. Repeated housing searches combined with interest in Boise employers may support an inference that someone is considering relocation even if the person never says so. A 2026 meta-analysis covering 79 studies and more than 51,000 participants found trust strongly associated with online disclosure and use behavior. The privacy calculus becomes more consequential when the benefit is no longer a better advertisement but delegation of work.

The Intent Economy and the Machine Customer

Businesses are beginning to serve a third participant alongside people and search engines: software representing the customer. An agent evaluating a hotel needs reliable information about availability and total price, along with clear cancellation terms and evidence of suitability. Human presentation still matters, but machine representatives require information that is structured, verifiable, and actionable.

The supporting ecosystem is already taking shape. By late 2025, Anthropic reported more than 10,000 active public Model Context Protocol servers. Google’s Agent2Agent initiative had support from more than 100 organizations, and its Universal Commerce Protocol launched with more than 20 commerce partners. Cloudflare also recorded a 305 percent increase in requests from OpenAI’s GPTBot between May 2024 and May 2025. Together, these developments suggest that the machine-facing Web is moving from experimentation toward practical deployment.

That shift changes Internet economics. The commercial Web competed for attention, while search engines gained power by standing between intention and destination. AI can move closer to intention itself. Pew found conventional-result clicks in about 15 percent of observed Google visits without an AI summary but 8 percent when one appeared; links inside summaries received clicks in only about 1 percent of visits.

The attention economy will persist, but an intent economy can form beside it. The original Web rewarded being a destination. Search rewarded being found. An agentic Web increasingly rewards being chosen. That also creates a financing problem. In June 2025, Cloudflare measured about 1,700 OpenAI crawls for every referral and about 73,000 for Anthropic. If machines consume more information while sending fewer humans to producers, the economics supporting the open Web may have to change.

Authority, Lock In, and the Invisible Web

Acting on intention requires more than identity. Authentication asks who the user is; agency asks what software may do for that user. Roughly 5 billion passkeys were active globally in 2026, while European digital-identity pilots involved more than 350 public and private organizations. Those developments strengthen authentication, but delegated authority remains a separate problem.

Consumers already draw the distinction. Roughly two-thirds use or would use AI shopping agents, yet nearly nine in ten want transparency and about half would stop using them if they lost control. Trust in independent AI agents stands at 28 percent. The likely path is staged: AI recommends, prepares, and eventually acts only within trusted and bounded limits.

Accumulated context may then become a new switching cost. A personal agent used for years could learn routines and preferences while incorporating relationships and thousands of corrections that make it progressively more useful. Agents could mediate $3 trillion to $5 trillion of global consumer commerce by 2030 under a moderate-adoption scenario. The five-year switching cost of personal context is not yet empirically measurable, so the lock-in argument remains a forward-looking inference. If learned preferences and delegated permissions cannot move between providers, the next platform lock-in may be the model of the user.

There will not be one Web of 2036. About 2.2 billion people remained offline in 2025, while 5G coverage was roughly 84 percent in high-income economies and 4 percent in low-income economies. Some populations may spend the next decade deciding how much authority to delegate to intelligent agents while others still confront basic barriers to reliable digital access.

Websites, search, screens, and human judgment will remain. What changes is the addition of systems capable of interpreting objectives and coordinating services while connecting information and performing some of the work once assigned to the person in front of the screen. The paradox is that the Web may become more important as it becomes less visible. For thirty years, humans learned how to navigate it. The next Web will increasingly navigate the digital world for us, raising a defining question: what must the Internet know about people to act for them, how much authority will they delegate, and who ultimately controls the intelligence standing between intention and the world?

Sources: FIDO Alliance, European Commission, Visa, McKinsey & Company, International Telecommunication Union

TL;DR Summary

• The Web is shifting from an information environment people navigate toward one that can increasingly interpret objectives and execute parts of tasks.
• AI agents may reduce search, coordination, communication, and transaction costs currently absorbed by users.
• User experience increasingly moves from navigating procedures toward expressing intention.
• Dynamic interfaces can reduce visible complexity even while fixed networks and data centers remain essential infrastructure.
• Connected systems make context more valuable than isolated behavioral records.
• AI inference changes privacy economics because systems can construct models from information users never explicitly disclose.
• Businesses increasingly need to serve both human customers and machine representatives.
• Agent-mediated selection can move economic power closer to user intention while weakening traditional referral economics.
• Open-Web financing becomes more difficult if machines consume information without generating comparable human traffic.
• Delegated agency requires authorization controls in addition to authentication.
• Accumulated personal context may create switching costs if learned preferences and permissions are not portable.
• The future Web will develop unevenly because infrastructure and access remain sharply unequal across economies.

Sources

• Pew Research Center; Americans and AI 2026 Chatbots Smart Devices and Views on Impact; – Link
• National Bureau of Economic Research; The Coasean Singularity Demand Supply and Market Design with AI Agents; – Link
• OECD; Artificial Intelligence and Competitive Dynamics in Downstream Markets; – Link

The Interface Stops Being the Destination

• International Telecommunication Union; Facts and Figures 2025 Internet Traffic; – Link
• Pew Research Center; Mobile Fact Sheet; – Link
• Google Developers; A2UI v0.9 The New Standard for Portable Framework Agnostic Generative UI; – Link

From Behavior to Context

• Pew Research Center; How Americans View AI and Its Impact on People and Society; – Link
• Pew Research Center; How Americans View Data Privacy; – Link
• Journal of Computer-Mediated Communication; Taking Stock of Social Media Privacy Meta Analysis; – Link
• World Wide Web Consortium; Digital Credentials API; – Link

The Intent Economy and the Machine Customer

• Pew Research Center; Google Users Are Less Likely to Click on Links When an AI Summary Appears; – Link
• Cloudflare; The Crawl to Click Gap Cloudflare Data on AI Bots Training and Referrals; – Link
• Anthropic; Donating the Model Context Protocol and Establishing the Agentic AI Foundation; – Link
• Google Developers; Universal Commerce Protocol; – Link

Authority Lock In and the Invisible Web

• FIDO Alliance; The State of Passkeys 2026 Global Consumer and Workforce Report; – Link
• Visa; Earning Consumer Trust in the Age of Agentic Commerce; – Link
• McKinsey & Company; The Agentic Commerce Opportunity; – Link
• International Telecommunication Union; Facts and Figures 2025 Mobile Network Coverage; – Link

Keywords: Internet Economics, Artificial Intelligence, User Experience, Agentic Web, Delegated Authority, Context Portability, Machine Intermediation