“The Internet Will Plunge into Darkness”: Newfangled AI Agents Have a Chance to Replace the Conventional Internet

Some analysts have called 2025 the year of AI agents – personalized digital assistants that can interact with users, conduct research, collect information, curate content, etc. As Bernstein analysts argue, “if AI agents actually become useful, the internet will go dark.”

Image source: Growtika/Unsplash.com

Websites and apps won’t disappear, but consumers won’t visit them directly because they’ll access information, content, and widgets through an AI assistant that will become an “aggregator of aggregators,” analysts say. If an AI agent can, for example, call a taxi, users won’t need to open an app to request a ride, they say.

AI agents representing users will become the new direct channel that tech companies use to connect with consumers. All other vendors will be routed through this new digital gateway and will likely have to pay some kind of fee — just as Google makes money from search ads and Apple makes money from its App Store commissions, Business Insider reports.

Big tech companies and startups are already battling for control of this promising segment of the AI ​​market. In late January, OpenAI unveiled an AI agent called Operator, which uses a web browser to perform actions on behalf of users, such as booking flights or buying groceries.

The key point is that the interaction now happens directly between the AI ​​agent and the user, whereas previously it had to be done through Google search. In the future, Google may be just one of many services available in the ecosystem of AI agents that OpenAI is developing.

Of course, Google has no intention of becoming just another app on a competitor’s platform. Last December, the company unveiled Project Mariner, an AI agent that can perform actions on the Internet on your behalf: browsing web pages, clicking buttons, and filling out forms. And earlier, in October, Anthropic unveiled a similar tool in test mode, allowing its Claude 3.5 Sonnet AI model to interact with any desktop application, simulating keystrokes, clicks, and mouse gestures—that is, to control a PC the same way humans do.

AI agents could also be controlled by voice, for example with smart glasses from Meta✴, or even mentally, using brain implants such as Elon Musk’s Neuralink.