The agentic web is here, reviving Semantic Web ideas

The agentic web is taking shape as AI systems begin to navigate websites, complete transactions and carry out tasks on behalf of users. The idea echoes Tim Berners-Lee's earlier Semantic Web vision of an internet designed for autonomous software agents, but today's large language models and tool-using agents are making it more practical. Instead of a person clicking through pages, an agent might compare products, book a service, fill out forms or coordinate with another agent. That shift could change how people search, shop and interact with websites. It could also reshape business models built on attention, advertising and direct user engagement. If agents do the browsing, companies may need to design for machine readability, clear permissions and reliable APIs. The agentic web raises hard questions about trust, safety, compatibility and who gets to decide how automated activity is allowed. How does a website know whether an agent is authorized to act? Who is responsible when an automated purchase goes wrong? How can users set limits on what an agent may share or do? The earlier Semantic Web struggled partly because it required widespread, careful markup that many publishers did not provide. Large language models lower that barrier by interpreting ordinary content and tools, but they also introduce new risks such as prompt injection, fake instructions and silent manipulation. The result may be a more automated internet that is convenient but harder to audit. Businesses will need to decide whether to welcome agents, block them or offer controlled access. Users will need clear ways to delegate authority and revoke it. Regulators may face pressure to define rules for agent identity and liability. The agentic web is not a distant concept. It is emerging through experiments and early products, and it is reviving old questions about how the internet should work when software, not people, is the main visitor.