Who's Your Agent? (doc.searls.com)

AI “agent” interfaces like OpenAI’s GPT Atlas and Perplexity’s Comet are forcing a rethink of what a browser is and who controls your online experience. Reporters note that some publishers now block Atlas from viewing their pages, returning messages such as “ChatGPT is unable to access the contents of this website.” Critics like Anil Dash call Atlas an “anti‑Web browser” because it often substitutes AI‑generated summaries or responses for the original site content, presents a command‑driven UX instead of clickable links, and effectively makes the human the agent for the browser rather than the browser acting for the user. The technical and ecosystem implications are big: agent layers can centralize content delivery, sidestep publisher monetization, and create a mismatch between what users expect (the live web) and what they get (model-generated approximations). Services like Comet emphasize delegation of tasks, but remain corporate‑hosted assistants with attendant privacy and ownership concerns despite FAQ reassurances. Expect a tug‑of‑war involving access control (publishers limiting crawlers or API access), legal and monetization disputes, UX design tradeoffs between human- and AI‑initiated interaction, and new standards for provenance so users can tell when they’re seeing the original web versus an AI’s rendition.