Gemini Can Now Use Your Mac for You: How to Try It Safely

Gemini chat on a Mac with the request ending in @Computer use, showing its steps and a live preview of Blender building a butterfly model Image: Google

Key takeaways

  • Google announced on Nov. 19 that the Gemini app for Mac can now use your computer for you, clicking, typing and filling out forms across desktop apps, local files and websites.
  • It works in a separate background window with a live preview in your chat, so your own screen and cursor stay free. You can pause or stop it at any time.
  • It’s rolling out in English in the U.S. to people 18 and older with personal Google accounts who pay for Google AI Plus, Pro or Ultra.
  • You need an Apple Silicon Mac running macOS Sequoia 15.0 or later. Intel Macs aren’t supported.
  • You pick which apps Gemini can use and approve sensitive steps, but Google itself warns about mistakes, deleted files and hidden “prompt injection” instructions. Keep it away from banking and passwords.

Google just gave its Gemini assistant hands. As of Thursday, Nov. 19, the Gemini app for Mac can take over the boring, click-heavy parts of a job: opening apps, moving through menus, typing information and filling in forms. Google calls the feature “computer use.”

That can save time on tasks you dread, but it also means an AI is acting inside your computer. Here’s what it does, who can get it, and how to use it safely.

What Gemini’s computer use actually does

In Google’s announcement, Gemini app product manager Michael Friedman explains that you describe what you need and Gemini “will navigate menus, type information, and click through apps to finish your work.”

The important detail is where it works. Gemini runs inside an isolated background window, and you see a live preview of each step right in your chat. Your own screen and cursor stay free while it works. TechCrunch notes that, like other AI agents with computer use, you can pause it, stop it and approve its actions before they happen.

Gemini's live preview on a Mac showing its steps in a chat and the Blender 3D app open in a small preview window as it builds a butterfly model
Gemini shows its steps and a live preview of the app it’s using, here Blender, inside the chat. Image: Google (frame from Google’s demo video)

Who can get it, and what it costs

This isn’t for everyone yet. According to Google and its Gemini release notes, you need:

  • To be in the U.S., using Gemini in English
  • A personal Google account, and to be 18 or older
  • A paid Google AI Plus, Pro or Ultra subscription
  • A Mac with Apple Silicon running macOS Sequoia 15.0 or later

Google’s U.S. subscriptions page currently lists Google AI Plus at $4.99 a month, Google AI Pro at $19.99 a month, and Google AI Ultra at $99.99 or $199.99 a month. The free Gemini plan doesn’t include computer use.

One wrinkle: AppleInsider spotted that Google’s support documentation currently mentions only Pro and Ultra for controlling desktop apps, even though the announcement includes Plus. Google also says the feature is “rolling out,” so it may take a few days to reach your account. TechCrunch reports Google didn’t say when it will expand to other countries or platforms.

How to get started, step by step

  1. Install the Gemini app for Mac. Go to gemini.google/mac and click Download for Mac. Open the .dmg file, drag the Gemini icon to your Applications folder, open it and sign in with your Google account, per Google’s help page.
  2. Update the app. Google notes some features may not work on an old version. In the Gemini app, click Gemini > Check for Updates at the top, then Install update if one is available.
  3. Grant Mac permissions when asked. AppleInsider reports that before Gemini can operate Mac apps, you have to enable computer use and give the app macOS Accessibility and Screen & System Audio Recording permissions. Gemini also asks before it controls each individual app.
  4. Ask for the job. 9to5Google says you add “@Computer use” to your prompt, and Google’s own demo image shows a request ending that way.
  5. Watch and approve. Follow the live preview, stop it with one click if it goes off track, and approve key actions when Gemini asks. Google’s release notes say Gemini pauses for logins or payments.

Good everyday uses

Google’s examples lean toward repetitive work that moves information from one place to another:

  • Forms from PDFs: pull weights, dimensions and addresses out of local PDFs to fill in a multi-box shipping quote.
  • Selling stuff you own: look at clothing photos in a folder, research prices, log item details and stage listings on resale sites.
  • Quick presentations: collect numbers and screenshots from online dashboards and build a Google Slides deck.

TechCrunch adds organizing files into folders based on rules you set and repeating the same step across a batch of files. One tip: “stage” is the key word in that resale example. Let Gemini prepare the listing, then read it over and publish it yourself.

Privacy and safety: set it up carefully

Google built in guardrails, but its own help pages are frank about the risks.

  • Give access app by app. AppleInsider notes that choosing “Always allow” lets Gemini use an app without asking each time. Save that for low-stakes apps. Don’t give it to your email, banking app or password manager.
  • Keep it away from money and logins. Google’s Mac task help page says not to type sign-in info, payment details or other sensitive info into the task thread. When a login or payment step comes up, take over and do it yourself. If you haven’t switched to passkeys yet, our guide to setting up passkeys on your phone and Google account will help.
  • Watch out for signed-in apps and codes. According to AppleInsider, Google warns that Gemini can reach active sessions in native Mac apps and can read or interact with two-factor codes shown in desktop or messaging apps.
  • Know about prompt injection. Google’s task help page explains that a website, email or document can hide instructions you can’t see but the AI can read, and those could steer it toward things you didn’t ask for, like sending your data somewhere. AppleInsider points out that untrusted content in one app could push Gemini to act in another app it has access to. Avoid pointing it at sketchy sites, the same ones we flag in our Black Friday scam red flags.
  • Back up first. Google says Gemini makes temporary backups when it edits your files, but they’re deleted when you start a new task or after 24 hours, and some changes can’t be undone. Don’t hand it files you can’t afford to lose.
  • Read before you approve. Gemini is designed to ask before deleting files or sharing them with other people. Check the files and the recipient every time.

AppleInsider’s take is the cautious one: avoid giving an AI agent broad control over your main Mac, especially if it holds sensitive files or accounts. Our guide on what to do after the Optimizely hack covers more account-safety basics.

How it compares

Google isn’t first here. TechCrunch notes that OpenAI, Anthropic and Meta already offer agent features in their desktop apps. On Nov. 17, Google also started rolling out Tasks in Gemini at no cost in the U.S. and a few other countries. Computer use on the Mac, though, needs a paid plan.

The bottom line

If you already pay for Google AI and own a recent Mac, computer use is worth a try for chores like copying PDF data into forms or sorting a messy Downloads folder. Start small, grant access app by app, and stay in charge of anything involving money, passwords or deleting files.

Sources

Leave a Reply

Your email address will not be published. Required fields are marked *