Summary of Key Points
Over the past 30 years, the fundamental premise of the internet has been to "show people content and provide them with services." All designs, such as navigation bars, buttons, and advertisements, have revolved around human needs. However, technologies like Google's Agent2Agent protocol (A2A) are beginning to challenge this premise: AI agents can collaborate in a standardized manner, and machines are becoming important participants on the internet. The internet is gradually dividing into two layers—the human layer (attractive pages, communities, stories) and the machine layer (structured data, protocols, AI collaboration). This not only changes the way we use the internet (from searching ourselves to having AI do it for us) but also redefines business logic (from competing for human attention to competing for machine trust), while introducing new challenges such as knowledge distortion and the governance of machine identities.
Detailed Analysis
1. The internet's "audience" has changed: from humans dominating to machines joining the game
The design of the internet was always focused on humans—navigation bars to help people find things, banners to catch their attention, and "Buy Now" buttons to encourage purchases—all of which are based on human needs. Machines, on the other hand, are interested in structured information: product specifications, inventory, return policies, payment methods, and the ability to execute transactions directly. For example, in a 2026 e-commerce experiment, the same website had two versions: one for humans and one "agent-ready" version that was easy for machines to understand and process. The result showed that AI had a 89.3% success rate on the machine-friendly version, compared to only 49.3% on the regular version. This means that websites in the future will need to answer two questions: "Does it look good to humans?" and "Can machines understand it?" This is not just about SEO optimization; AI may be able to complete transactions without even bringing users to the page.
2. A revolution in how we use the internet: from doing everything ourselves to having AI do it for us
The traditional process of using the internet involved identifying a need, searching, opening multiple pages, comparing options, and placing orders—all done by humans. Now, AI agents can streamline this process with just one command, such as "Buy hiking shoes for $150 that will be delivered by Friday." The rest of the tasks—searching, comparing prices, checking logistics, and making payments—are handled by AI in the background. This is not just a concept; services like OpenAI's Instant Checkout and Stripe's integration with ChatGPT allow users to complete purchases directly through chat. Essentially, we are delegating the process of using the internet to AI. Businesses are now dealing with AI agents representing customers, an audience that has never existed before.
3. Business logic is being redefined: from competing for human attention to competing for machine trust
The core of internet commerce used to be about capturing human attention—page views, time spent on a site, and ad impressions. However, machines don't have attention; they are not attracted by attractive pages or impulsive purchases driven by celebrity endorsements. They care only about whether the data is clear, the service is reliable, and whether it can be easily used. Adobe predicts that future marketing will involve three types of relationships: human-human, human-AI, and AI-AI. In an AI-AI world, the most valuable brands will be those that are useful: providing verifiable data, easy-to-use services, and reliable after-sales support. In the future, a visually appealing website may lose to one with clear and structured information because AI will prefer the latter.
4. Hidden risks: The cycle of machines leading to knowledge distortion
When machines become the primary users of the internet, content enters a loop of "machine-generated → machine-read → machine-generated again." For instance, a 200-page technical document might be created and processed entirely by AI without anyone reading it. Over time, this could lead to knowledge echoes: machines referencing each other, with information drifting away from its original source, and ultimately, no one knowing the truth. This is more concerning than fake news, as fake news can be traced to its creators, but in the machine cycle, each AI is simply summarizing content faithfully, without lying, yet still distorting the facts.
5. New challenges in governance: Giving machines "identities"
The management of web crawlers used to be general; now, it's necessary to distinguish between different types of AI traffic. For example, Cloudflare categorizes AI traffic into three types: Search (crawlers for searching), Agent (AI agents commissioned by users), and Training (crawlers for model training) and allows websites to set rules accordingly—pages with ads block Training and Agent traffic by default because ads are for humans. In the future, AI agents will need "identities" that specify who is accessing them, what they are doing, and how long they can stay on a site, similar to how web browsers identify users. This involves legal and regulatory considerations: "malicious robots" and "user-commissioned AI" must not be confused, and the rights of agents should reflect those of the people they represent.
Conclusion
The internet will not disappear, but for the first time, its primary audience will no longer be solely human. In the future, we will have attractive human-facing interfaces, while most business activities will occur in the invisible machine layer. The standard for measuring the value of content will shift from "how many people have seen it" to "how many machines have read it, believed in it, and acted on it." This represents the most profound transformation the internet has undergone in the past 30 years.