Gemini can now use your Mac for you: what it does, who gets it and what to watch

Google is adding computer use to the Gemini app for Mac: the AI clicks, types and fills in forms in your apps while you keep working. For now it's US-only, English-only and for paid subscribers.

A desk with a laptop, a phone and a person taking notes
Image: Firmbee en Pixabay

Until now, Gemini could tell you how to do things on your computer. Since November 19, the Mac app can do them itself: open an app, click through menus, type and fill in forms. Google calls it computer use and announced it on its official blog.

It’s another step toward AI agents: assistants that don’t just answer, but act. For the bigger picture and the risks, see our guide on what an AI agent is.

At a glance

What it is Gemini operates apps and websites on your Mac to complete multi-step tasks
Where US only, English only, for now
Who Google AI Plus, Pro or Ultra subscribers (not the free version)
Requirements Apple silicon Mac (M1 or later) on macOS Sequoia 15 or newer; personal account, 18+
Other countries? Google hasn’t said when
Windows? Not announced

What it can do

You describe what you want and Gemini clicks and types for you. Google’s own examples:

  • Color-grading a batch of photos in design software.
  • Pulling weights, sizes and addresses out of several PDFs into a shipping quote form.
  • Gathering numbers from dashboards and building a Google Slides deck.
  • Looking at clothing photos in a folder, researching prices and drafting resale listings.

TechCrunch adds simpler uses, like sorting files into folders or repeating the same step across many files.

How it works

  1. Download the Gemini app for Mac (gemini.google/mac).
  2. Describe the task. According to 9to5Google, you trigger it by adding “@Computer use” to your prompt.
  3. Choose which apps it may use.
  4. Gemini works in a separate background window, with a live preview in the chat. Your screen and cursor stay free.

Is it safe to let an AI use my computer?

Google lists several controls:

  • You can watch every step live.
  • You can pause or stop it at any time.
  • It asks you to approve key actions before they happen.
  • You decide which apps it can touch.

Still, be careful. An agent can get things wrong just like a chatbot (a so-called hallucination), but here a mistake isn’t a bad sentence: it’s a moved file or a form sent with wrong data. Our advice:

  • Start with tasks you can undo (work on copies, not originals).
  • Don’t give it access to banking, work email or apps with sensitive data unless you need to.
  • Check the result before anything gets sent.

A bit late to the party

TechCrunch notes that OpenAI, Anthropic and Meta already offer similar features and calls Google “a little late to the game.” What’s different is that it lives inside the Gemini desktop app, next to Gemini Tasks, which already make calls and purchases in some countries.

Bottom line

Useful for repetitive chores, but for now only for paying Google AI subscribers in the US on a Mac. We’ll update this when it expands.

Sources