Review of 9 AI Chatbots Across 8 Browsers Finds They Reshape Web Use

3 articles · Updated · The New York Times · May 18

Nine AI chatbots tested across eight browsers led the reviewer to conclude that built-in assistants can materially change daily web use by keeping context inside the page and reducing tab-switching.
Months of use showed the tools work best for synthesis, writing and research—such as planning trips, comparing prices and generating reading lists—while the reviewer said brand-specific extras mattered less than using a familiar browser or chatbot.
Agentic features still lagged: browsers struggled to secure a hard-to-get dinner reservation, surfaced obvious options and needed repeated help with forms, often making manual booking faster.
Security limits help explain that gap. University of Toronto professor Colin Raffel said browsers must sandbox AI, because agents with access to sensitive data, the internet and autonomous actions can create phishing, prompt-injection and spending risks.
The article’s takeaway is pragmatic: AI browsers are already useful as an always-available research aide, but users should avoid exposing sensitive accounts, verify outputs and be cautious about letting bots act on their behalf.

As AI browsers become more powerful, are we trading our online security for simple convenience?

Experts call today's AI agents 'middle schoolers.' What will it take for them to finally grow up?

AI can find any clue in a treasure hunt, so why do we still prefer human partners?