Say you have a folder of clothing photos you want to sell online. Google’s pitch is that Gemini on your Mac studies the photos, researches prices, logs each item and stages the listings while you keep working. That resale chore is one of the examples Google gave when it added computer use to the Gemini app for Mac on 19 November.

Computer use means the AI operates software the way you would: it clicks, types, scrolls and works through menus. Gemini does this in a separate virtual screen, so your cursor stays free, and a live preview in the chat shows what it is doing.

What people can hand off

Google’s other examples are practical. Batch color grade photos in design software. Pull weights, dimensions and addresses from local PDFs into a shipping quote. Gather dashboard metrics and screenshots into a Google Slides deck.

The Gemini help page for Mac apps adds expense reports filled from receipts in your Downloads folder. Another example imports trip photos into Lightroom, applies one lighting filter to all of them and exports them as 4K PNG files. 9to5Google spotted new desktop connectors for apps such as Blender and Autodesk Fusion.

New to the idea? Our plain guide to how AI agents differ from chatbots covers the basics.

How to try it

You need an Apple Silicon Mac on macOS Sequoia 15 or later, a personal Google account and to be 18 or over. For now it works only in English in the US, and not on work or school accounts.

In Gemini Settings, open Computer use and switch it on, then give macOS the Accessibility and screen recording permissions it asks for. Click Task under the chat box, type @Computer use and describe the job. Google recommends Gemini Hydrogen 4 or a newer model.

What it costs

It comes with Google AI Plus, Pro and Ultra. Google’s AI plans page lists prices of $4.99 a month for Plus, $19.99 for Pro, and $99.99 or $199.99 for the two Ultra tiers. Higher plans get more use of Tasks in Gemini, but the page gives no exact limits.

AppleInsider noticed the help page first named only Pro and Ultra. When we checked it on 20 November, it listed Plus too.

Where it can go wrong

Google’s help page is unusually blunt. Gemini works inside apps where you are already signed in, and it can read login codes that appear in messaging apps. Content in one app can also steer it into acting in another app you allowed. That is prompt injection: hidden instructions the AI follows as if you had typed them.

Access to a code editor or developer tool can lead to shell commands and wider file changes. Gemini tries to back up files before editing and keeps copies for 24 hours or until your next prompt. Some apps cannot be backed up, though, and some actions cannot be undone.

There are guardrails. Gemini asks the first time it wants an app and again before sensitive steps, and 9to5Google says it pauses for logins and payments. That resembles the per-purchase check in ChatGPT Wallet’s agent shopping. Choosing “always allow” removes the app prompt from then on.

Apple is tightening its side too. The Verge reports that a macOS 27.3 beta previews new limits on apps that want full disk access. We have not tried Gemini’s computer use ourselves, so the examples here are Google’s.

What to do this week

On a US Plus, Pro or Ultra plan, start with one low-stakes chore, such as filtering a copy of a photo folder. Turn on file backups and pick “allow once” over “always allow”. Put your password manager, email and messaging apps on the blocked list, and keep code editors off the allowed one.

Sources

Anna Jakovleva

Anna Jakovleva is the senior editor for AI tools at AI Observer, leading coverage of agents and automation, design and product, and the site's tutorials. She worked in operations and product teams before joining, so she judges tools by setup time, cost and where they break. Her writing is hands-on and practical.

Spotted a mistake? Tell us on the corrections page.

Conversation

Every comment is reviewed before it appears, so there can be a short delay.