Google’s new Gemini AI can actually use the web like you do

Just a heads up, if you buy something through our links, we may get a small share of the sale. It’s one of the ways we keep the lights on here. Click here for more.

Google’s latest AI experiment might just make your browser feel like it’s haunted, in a good way. 

The company has unveiled Gemini 2.5 Computer Use, a new model that can actually use a web browser like a person would. 

Instead of relying on APIs or hidden shortcuts, it clicks buttons, fills out forms, and drags stuff around, all by looking at what’s on screen. 

Think of it as a digital intern who actually reads the page before breaking something.

The new model uses what Google calls “visual understanding and reasoning” to interpret what it sees and complete tasks based on a user’s request. 

So, if you ask it to fill out a form, it won’t just send data. It will literally find the right boxes and type in your info like a human. 

Google says it’s handy for UI testing or for websites that don’t offer direct access through an API. It’s AI that can play nice with human-designed interfaces.

If that sounds familiar, it’s because this “agentic AI” arms race is heating up fast. 

Just yesterday, OpenAI showed off new ChatGPT apps and its upcoming ChatGPT Agent, which can perform tasks automatically. 

Anthropic also rolled out a “computer use” feature for its Claude AI last year. 

Now Google’s jumping in, saying its model “outperforms leading alternatives” on web and mobile benchmarks. Though, to be fair, its demos are sped up 3x, so take that with a grain of algorithmic salt.

Unlike OpenAI’s approach, Gemini 2.5 doesn’t yet control your whole computer. It’s limited to a browser sandbox for now, supporting 13 actions like typing, scrolling, and dragging items around. 

Still, that’s enough for it to do things like play the game 2048 or dig through Hacker News for spicy debates.

Developers can already try it out through Google AI Studio or Vertex AI, and there’s even a public demo on Browserbase where you can watch the AI in action.