Making the internet better for AI agents might make it worse for everyone else (www.technologyreview.com)

The rapidly growing use of personal AI agents, such as OpenAI’s Dots and Meta’s Muse, is reshaping the internet landscape, introducing a third category of web users that complicates existing frameworks designed to distinguish between bots and humans. These AI agents can perform tasks online at machine speed on behalf of individual users, resulting in a surge of agentic web traffic. However, their integration has led to increased website restrictions, with many platforms blocking AI agents entirely to protect their user interfaces and business models, particularly in e-commerce. This shift has significant implications for the AI/ML community and internet users at large. As demand for AI agents rises, platforms may adapt, but the adjustment could lead to an internet increasingly optimized for these agents, potentially degrading the experience for human users. Websites may become less navigable for people, with challenging verification processes and differential pricing based on whether an interaction is handled by an AI agent. The outcome largely rests on how businesses set rules for agent interactions, balancing efficiency and security while determining the future role of AI in consumer negotiations and everyday online tasks.