Google has finally unveiled a secretive AI tool that can take control of a person’s browser in order to navigate the web in the same way as a human – offering a fundamental shift in the way we interact with the internet.
Project Mariner, which was previously known internally as ‘Project Jarvis’, could represent the biggest update to web browsing since its inception 34 years ago. It also forms an integral part of the US tech giant’s mission to create a universal AI assistant that can carry out everyday tasks – from booking holidays, to paying for parking tickets.
Still in its early stages, Project Mariner comes in the form of an experimental extension for Google’s hugely popular Chrome web browser. It uses the firm’s Gemini AI to understand the information on a person’s screen in order to take the necessary actions.
By analysing pixels, images, text and web forms, the AI is able to complete tasks by typing or clicking in the appropriate places – though not always very well. It is currently only available to a limited number of testers, and Google admits that the current version is much slower than a human and still not as reliable.
“It’s still early, but Project Mariner shows that it’s becoming technically possible to navigate within a browser, even though it’s not always accurate and slow to complete tasks today, which will improve rapidly over time,” Google wrote in a blog post introducing the tool.
Google describes Project Mariner as a paradigm shift in user experience, which will fundamentally change the way we interact with websites and technology more generally.
If successful, it could also force businesses to shift the way they operate in order to better accommodate AI agents. Entire industries, such as advertising, could also be disrupted, as humans will no longer be visiting websites to view online ads.
Project Mariner forms part of a major update to Google’s artificial intelligence offerings, which centres around the next-generation of its AI virtual assistant Gemini. Since its launch last December, Gemini has been designed to be multimodal, meaning it can understand information via text, audio, video and image inputs. The latest version is also able to produce information through these outputs.
Gemini 2.0 comes with a number of prototype AI agents that are designed to be experts at specific tasks, such as planning holidays or writing code.
Google chief executive Sundar Pichai said the latest advancements represent what he refers to as AI’s “agentic era”, which forms part of the company’s vision of creating a universal assistant.
“We have been investing in developing more agentic models, meaning they can understand more about the world around you, think multiple steps ahead, and take action on your behalf, with your supervision,” Pichai said. “We continue to push the frontiers of AI to organise the world’s information across every input and make it accessible via any output, so that it can be truly useful for you.”
Other agents unveiled on Wednesday include Jules, which helps developers with coding tasks, and Deep Research, which uses advanced reasoning to act as a research assistant.
