Khaberni - Google has endowed its artificial intelligence tool "Gemini" with capabilities that mimic AI clients through its own browser that can visit websites and interact with them as humans do, according to a report published by the tech site "The Verge".
The new capabilities come as part of the "Gemini 2.5 Computer Use" model showcased by the company in an official blog post with several videos demonstrating the model's various uses.
The new model operates similarly to conventional AI agents, interacting directly with websites through its own interface instead of relying on a customized programming interface to leverage the features of other tools and sites.
The company previously employed this model experimentally in several features, including placing AI in search engines as well as "Project Mariner," a test project that enabled AI tools to add products to a shopping cart based on a provided list.
The new model can be used in many aspects, like testing website interfaces to ensure they all function well or filling out surveys and scanning websites for information.
The revelation of this new model coincides with "OpenAI" introducing the "ChatGPT Apps" feature, which allows the use of a set of applications directly within "ChatGPT" without the need to leave the tool's window.
Meanwhile, "Anthropic" launched a similar tool to the "Cloud" AI model last year, and Google asserts that its tool exceeds competitors in performance evaluation tests on computers and phones.
Unlike competitors, "Gemini" does not have an integrated computer it operates through, meaning this new feature is limited to the internet browser and the sites that can be visited through it.
However, it avoids the need for direct programming linkage with the APIs of different services, meaning you can use the site without needing to link the tool to it.
The new tool is still experimental and primarily available to developers, and it is expected that Google will soon release it to the rest of the users.
