Skip to content

Project Mariner

An early research prototype that acts inside the browser

Visit official site

What Project Mariner is

The project is described as an early research prototype from Google DeepMind based on Gemini 2.0, exploring future human-computer interaction inside the browser.

What you can do with it

  • Read text, code, images and forms on a screen
  • Carry out tasks through a Chrome extension
  • Fill forms and search for information
  • Request confirmation before sensitive actions

Application scenarios

  • Finish shopping or lookup tasks without manual steps
  • Bring the API into another application
  • Analyse interaction data for product design

What to watch out for

  • The prototype is described as early research
  • The extension is described as experimental
  • Sensitive operations ask for final user confirmation

Pros & cons

✓ What we like

  • Sensitive actions request final user confirmation
  • Interaction with the web ecosystem is described as ongoing learning

! What to watch out for

  • The material describes it as an early research prototype
  • Task execution runs through an experimental extension

FAQ

Which model is it built on?

The material states it is built on the Gemini 2.0 model.

What benchmark result is named?

The material states an 83.5 percent completion rate on WebVoyager.

Last reviewed: 2026-11-15

More AI search engine tools

View all →

How we review