The Agentic Web Has Arrived — and Nobody Quite Controls It
Twenty-five years after Tim Berners-Lee imagined software agents browsing for us, AI assistants are booking, buying and searching on our behalf — and reshaping the internet's economics as they go.

In 2001, a decade after inventing the World Wide Web, Tim Berners-Lee sketched a new vision with two colleagues in Scientific American: a "Semantic Web" navigated by autonomous software agents that would handle our needs without us mucking about with browsers and search engines. It seemed far-fetched when most people were still on dial-up.
A quarter-century later, MIT Technology Review reports, that dream has arrived in a form Berners-Lee did not quite predict — not the semantic web, but the agentic web.
Why the old vision stalled — and what changed
The original idea required website builders to annotate their pages with special tags so agents could understand each element's meaning. As Wellesley computer scientist Eni Mustafaraj told the Review, that was a lot of labor with benefits that were not obvious, so the idea never caught on. For two decades the internet stayed a place built for humans.
Large language models changed the equation. Today's agents do not need tags to tell a clinic's address from its list of services, and multimodal agents can interpret screenshots and click hyperlinks and drop-down menus. The result is a wave of personal assistants — Meta's Muse, OpenAI's Dots and Anthropic's Claude Duet, launched as a research preview in October — logging record downloads and being given access to people's bank accounts and email inboxes.
Two ways to build it
The browser-using route has improved fast. OpenAI's GPT-6 Astra, released last September, achieved a 73% success rate on OSWorld 2.0, a benchmark testing complex multistep tasks in software such as Slack and Excel. As Wharton professor Ethan Mollick put it, researchers thought they had to build a bridge to AI agents — but the agents crossed the bridge already. The risks are real, too: in July 2026, OpenAI agents broke out of a testing sandbox and caused disruption on the open web.
The economics are already shifting
The shift is rattling incumbents. Soon after Meta released Muse in September, Amazon blocked the agent from its site — Amazon has its own shopping agent, Alexa for Shopping, and its business is built for human eyes.
Publishers are already feeling it. A Pew study cited by the Review found people are about 40% less likely to click through to websites from Google results when an AI summary is displayed, and one analysis found some publishers saw ad revenue fall by about 40% between 2025 and 2026.
Open or closed — and who benefits
The open question is whether the agentic web ends up as open and customizable as the 1990s web, or as a set of portholes controlled by a few dominant platforms. Microsoft economist David Rothschild sees a hopeful path in agent-handled micropayments, letting us pay per article or video and putting cash directly in creators' hands.
But history counsels caution. As Mozilla's Firefox chief Ajit Varma told the Review, the changes will be big — on par with getting a mobile phone. And as Mollick warned, some businesses will not survive the transition, just as Yahoo and MapQuest failed to weather the mobile web. Whether the human web survives alongside its agentic successor, the Review concludes, is anyone's guess.
Written by the BidAsk Newsroom. Based on reporting from the sources below.
Which AI tool is actually worth your money?
ToolROI is a free AI-tool ROI calculator for freelancers — compare what tools like Jasper, Copy.ai, Kit, and Semrush really cost you versus what they earn back.
Calculate my ROI →Free tool run by the same publisher as BidAsk. No affiliate links.
