Search engines are becoming answer engines, websites are becoming data sources and advertisements are now being written for machines. The result is a new web in which AI platforms increasingly own the audience, the recommendation and the transaction.

It’s increasingly hard to recognize the World Wide Web. It’s become the wild Wild West, a brave New World, which brings with it both opportunity, risk, and for some businesses, catastrophic change.

Many of these changes are a result of two factors: AI and Google stranglehold on search, which is exerting itself in ever more unanticipated and negative impacts on both user experience and business customers.

For most of the commercial web’s history, the arrangement was relatively simple. Publishers and businesses made their pages available to search engines. Search engines indexed those pages and sent people back through links. Website owners converted that traffic into advertising revenue, subscriptions, leads or sales.

That bargain is breaking. In an article provocatively titled “The Week the Open Web Died,” Marketecture founder Ari Paparo describes a market moving through three overlapping phases: “AI in Ads, Ads in AI, Ads for AI.”

The third category is the most novel, and highly revealing. Advertisements are no longer being created solely for people. Some are being written specifically for AI reading, with the intent that the information contained in them will influence what an AI says about a company later.

This all doesn’t mean the technical web is going away. But it is changing. Websites, URLs and open internet protocols remain. What is shrinking is the open web’s role as the place where people discover information, consume content and encounter advertising. This is not the first major change that the Web has experienced, which occurred with the introduction of the mobile era, social networks and apps. Many things changed then, but what changed was the presentation layer, not the audience. This is different.

The mode of interaction is shifting. The web is increasingly becoming the wholesale supplier of information. AI platforms are becoming the point of interaction and storefront. There are now essentially three overlapping ways of accessing online information through the WWW: through walled gardens, where platforms control access and distribution; through the open web, where people navigate public, linkable sites; and through the machine web, where AI systems retrieve, interpret and increasingly act on information for people.

The referral bargain is collapsing

The decline in open-web traffic is measurable and indisputable. The Reuters Institute found that Google organic search traffic to more than 2,500 news sites fell 33% globally and 38% in the United States between November 2024 and November 2025. Media executives surveyed by the institute expect search referrals to decline another 43% over the next three years.

AI-generated search results help explain why. A Pew Research Center study found that users clicked a conventional search result in 15% of visits when no AI summary appeared. When an AI summary was present, that fell to 8%. Only 1% of visits resulted in a click on one of the sources cited inside the AI summary.

The economic effects are already appearing. Ozone data covering approximately 20 billion impressions found that publisher ad-request volumes in the second quarter of 2026 were down 32% to 37% year over year in the United States and 39% to 41% in the United Kingdom.

That figure measures advertising opportunities rather than total web traffic, but the relationship is direct: fewer visits produce fewer page views, fewer ad calls and less inventory for publishers to sell.

Search engines once gave users a map. Answer engines increasingly give them a destination. The website that supplied the information may receive a citation, but it does not necessarily receive the visitor.

The first advertisements for machines

TIME has taken the next logical step in this transition.

The publisher created stripped-down markdown versions of its pages that are easier for AI crawlers to process. Working with ad-tech company Mobian, it then began placing advertisements inside those machine-readable pages.

The initial advertisers included Ally Bank and the Project Management Institute. Their messages were converted into FAQ-style text, identified as sponsored content and placed where AI agents could retrieve them. TIME’s chief operating officer told Digiday that the publication now sees more bot traffic than human traffic on most days.

These are fundamentally different from ads shown inside an AI chatbot. An ad inside ChatGPT, Google or another AI product is still presented to a person. An advertisement for an AI agent attempts to influence the information the machine uses when forming an answer. The advertisement becomes a source.

If the strategy works, an AI might later describe, recommend or compare the advertiser using information it encountered in that paid placement. The human receiving the answer may never see the original advertisement or know that part of the answer originated in sponsored material.

The experiment immediately exposed the trust problem. On August 11, Perplexity confirmed that it had blocked TIME’s markdown ads from influencing its search index. It also warned that publishers using such practices could receive lower trust scores.

TIME had labelled the material as sponsored. Perplexity still objected because the label could disappear when an AI extracted a claim and reproduced it as part of an answer.

That dispute will not end advertising for agents. It establishes the battle that will shape it: who decides whether machine-readable information is evidence, advertising, manipulation or some combination of all three?

From a web of pages to a web of answers and actions

The web’s emerging structure looks substantially different from the one marketers learned to use.

The likely destination is a two-layer web. The human-facing layer will become more direct and deliberate: subscriptions, apps, newsletters, communities, events, customer portals and other environments where companies control the relationship.

The machine-facing layer will contain structured product information, corporate facts, prices, availability, policies, research, reviews and transaction capabilities. AI systems will retrieve this material through webpages, feeds, APIs and emerging agent protocols.

Access to that layer will not always be free. Cloudflare now allows website owners to distinguish among search, agent and training bots, and it is developing infrastructure for paid, programmatic access to web resources. The future open web may consequently be more permissioned, metered and commercially negotiated than the one it replaces.

Above both layers will sit the AI interface, the increasingly closed environment where the user asks the question, receives the recommendation and may complete the transaction.

What this means for search

SEO remains relevant because answer engines still need information to retrieve. Google’s own guidance says pages must be indexed and eligible to appear in conventional search before they can appear as supporting sources in AI Overviews or AI Mode. It also says no special markdown, AI markup or llms.txt file is required. Conventional technical SEO, original information and reliable, people-first content remain the foundation.

The objective of search strategy is expanding, however. Companies must care about whether an AI can find their information, whether it interprets that information correctly, whether it includes the company in a recommended shortlist and how it describes the company relative to competitors.

Rankings and clicks remain useful metrics. They now sit beside:

• Citation frequency
• Share of AI recommendations
• Accuracy of generated brand descriptions
• Inclusion in product or vendor comparisons
• Referral quality
• Agent-assisted leads and transactions

A company may lose traffic while gaining influence. It may also appear frequently in AI answers without receiving enough attribution or revenue to justify producing the information being used.

What business owners should do

The rise of answer engines makes the company website more important as a source of accurate, updated data, even when fewer people visit it directly.

Product details, services, locations, pricing, availability, credentials, policies and contact information should be current, consistent and easy for both humans and machines to understand. Important information should not be buried exclusively in images, PDFs or complicated interactive interfaces.

Businesses should also review which automated systems can access their sites. OpenAI, for example, provides separate controls for its search crawler and training crawler. A site can allow OAI-SearchBot while blocking GPTBot, preserving eligibility for ChatGPT search without granting the same permission for model training.

Owners should avoid treating machine optimization as a collection of tricks. A separate AI-facing page containing claims that customers cannot see may be interpreted as cloaking or manipulation. Different formatting may be useful; different facts are dangerous.

The durable advantage will come from publishing information an AI cannot obtain from thousands of interchangeable sites: original research, documented experience, named expertise, strong customer evidence and frequently updated first-party data.

Businesses also need direct audience channels. An email list, customer account, membership, app or community provides a relationship that cannot disappear with a search-interface update.

What corporate marketers should do

Corporate marketing is beginning to acquire a new responsibility: managing what machines know about the company.

That requires closer coordination among marketing, communications, product, ecommerce, data, legal and investor relations. Contradictory product descriptions, outdated corporate pages and inconsistent pricing are no longer minor website problems. They can become conflicting evidence inside AI-generated answers.

Marketers should regularly test the questions customers ask AI systems:

• Which companies does the system recommend?
• How does it describe the brand?
• Which sources does it cite?
• Which facts are missing or wrong?
• Does it understand the company’s products, customers and competitive position?
• Can it retrieve current price, inventory or availability information?
• Can an agent complete the next step successfully?

Paid opportunities inside AI products will grow, as will experiments aimed directly at agents. Corporate marketers should test them carefully while demanding clear separation between organic answers and sponsored information, reliable provenance and evidence that a paid message produced a measurable business result.

The TIME experiment demonstrates the risk of moving too aggressively. A campaign designed to improve AI visibility could instead cause an answer engine to lower the publisher’s or advertiser’s trust.

In this environment, trust is part of technical performance.

The web after the open web

The open web still contains the original reporting, product information, expertise and evidence that AI systems require. That gives it enduring value. Ozone’s data even found that prices increased for some of the remaining premium advertising inventory as supply contracted. Its position in the value chain is changing.

Websites are moving from the front of the customer experience to the infrastructure underneath it. AI platforms increasingly control discovery, synthesis, recommendation and action. Publishers and businesses supply the facts while the platform decides which facts are visible, how they are framed and whether the original source receives a visitor.

The open web’s future therefore depends on establishing a new exchange of value. That could include referrals, prominent citations, licensing, paid crawler access, direct transactions or some combination of them.

Without that exchange, more publishers will close their doors to machines, more information will move behind commercial agreements and a small number of AI platforms will become the arbiters of what consumers know about companies.

The web is not running out of information. There is more content than ever, and with the advent of AI the ease of creating websites increases. But who reads that information, who packages it and who gets paid when it produces value, is changing in fundamental and likely irreversible ways. Things made for humans are now being made for machines, agentic enabled machines that can and do make decisions you may not even know are happening.