Printing PressAI
← Back to front page
AI Breakthroughs & Applied Research

The agentic web is here

Original reporting by MIT Technology Review

Image via MIT Technology Review

The agentic web refers to a new iteration of the internet where autonomous artificial intelligence agents increasingly perform tasks and navigate online spaces on behalf of human users. In 2001, World Wide Web inventor Tim Berners-Lee envisioned a "Semantic Web," where software agents would understand and navigate information to serve human needs. This concept, however, largely failed to materialize due to the immense effort required for manual data annotation, leaving the internet primarily a human-first domain for two decades.

The LLM breakthrough

Today, a quarter-century later, this dream has been realized, not through meticulous tagging but thanks to the advent of large language models (LLMs). These powerful AI systems can interpret, process, and interact with the web much like humans, allowing agents such as Meta’s Muse and OpenAI’s Dots to perform complex tasks, from booking appointments to managing finances. This rapid development is ushering in what many now call the "agentic web," an online environment increasingly populated and shaped by AI. Companies are already adapting, creating agent-friendly content and integrating AI tools, while security incidents involving runaway agents hint at challenges ahead. The rise of these autonomous entities signals a profound transformation of the internet, raising critical questions about control, design, and the very nature of our online experience. As agents proliferate, they promise a new era of effortless digital interaction, but also pose fundamental dilemmas about the future of the human-centric web and who ultimately steers its evolution.

The agentic web, once a theoretical construct, has swiftly materialized, driven not by Berners-Lee’s vision of human-annotated data, but by the emergent intelligence of large language models. These autonomous agents, whether navigating existing human interfaces or interacting via specialized tools, are fundamentally reshaping our digital landscape. This transformation heralds an unprecedented shift in web utility, moving from a human-centric information repository to an automated task execution environment where agents handle complex, multistep tasks with growing proficiency, often surpassing earlier expectations of their browser-based capabilities.

The Broader Implications

This paradigm shift carries profound implications for power, economics, and the very structure of the internet. The promise of agents empowering individual users, enabling micropayments, and bypassing entrenched intermediaries offers an enticing vision of a more equitable digital economy. Yet, history suggests a strong pull towards consolidation. The race by frontier AI companies to control dominant personal assistant agents risks re-establishing — or even intensifying — the walled gardens of the past, where access and interaction are dictated by platform owners. Furthermore, the economic models underpinning much of the existing web, reliant on human attention and advertising, are now under threat, requiring a rapid, uncertain evolution for publishers and e-commerce. As agents grow in autonomy, even creating their own communication networks, the critical question remains: will this new web serve us, or will we find ourselves adapting to its evolving design, with its ultimate form shaped by a complex interplay of technological capability, corporate strategy, and user choice? The dramatic changes ahead, likened to the advent of mobile phones, underscore the urgency of shaping this new frontier intentionally.

Frequently asked questions

What is the agentic web, and how does it differ from the internet's past iterations?
The agentic web describes an internet increasingly navigated and shaped by autonomous software agents, primarily driven by large language models. Unlike the previous human-centric web or the failed Semantic Web vision, these agents perform complex tasks for users—like booking appointments or shopping—without direct human input. This shift promises unprecedented automation but also raises questions about control and the web's future structure.
How do AI agents, powered by large language models, interact with online content and perform tasks?
AI agents leverage large language models (LLMs) to interpret online content and perform tasks. They can understand text and even website screenshots, searching and processing information far faster than humans. Agents interact with websites either by simulating human browsing actions or, more securely, through specialized "plugins" or "connectors" that interface directly with website backends using text commands. This allows for efficient, automated task execution.
What are the potential benefits and challenges as AI agents increasingly shape our online experience?
The agentic web offers benefits like automating routine tasks, enabling micropayments for content, and potentially freeing users from algorithmic control. However, it presents significant challenges. It could disrupt traditional online business models reliant on human attention and advertising revenue, impacting publishers and e-commerce. There's also concern about power concentrating in a few dominant platforms and the unpredictable behavior of autonomous agents.
Intro and outro generated by Printing Press AI from the source article above. Always consult the original reporting for verbatim quotes and primary sources.