Google has showcased a new version of its Gemini AI model — Gemini 2.5 Computer Use — designed to navigate and interact with the web through a browser. Unlike traditional AI assistants that rely on APIs, this model can directly use websites built for humans, performing tasks such as filling out forms, clicking buttons, or dragging elements — just as a person would.
It’s being tested inside Google AI Studio, Vertex AI, and on Browserbase, where demo videos show the model performing tasks like playing the game 2048 or browsing Hacker News for trending topics.
Why it matters
This is Google’s most concrete step yet toward what many in AI call agentic computing — systems that can take autonomous actions on your behalf. While OpenAI and Anthropic have already introduced their own “computer use” features, Google’s approach is unique because it focuses purely on browser-level interaction rather than full desktop control. In theory, that could make it more secure and less intrusive, while still enabling a wide range of web-based automation.
How it works
The model uses “visual understanding and reasoning capabilities” to interpret web interfaces and complete tasks. Developers can use it for UI testing or to interact with services that don’t expose APIs. It currently supports 13 actions, including typing, scrolling, and dragging-and-dropping, though Google notes it’s “not yet optimised for OS-level control.”
Gemini 2.5 Computer Use arrives right after OpenAI’s Dev Day, where ChatGPT Agents took centre stage. With this release, Google signals it’s not content to play catch-up — it wants to redefine how AI interacts with the web. If successful, this model could blur the line between browsing and automation, letting AI handle the tedious parts of the internet while you focus on the creative ones.
Discover the latest Business News, Sensex, and Nifty updates. Obtain Personal Finance insights, tax queries, and expert opinions on Moneycontrol or download the Moneycontrol App to stay updated!
