Google began rolling out “computer use” in the Gemini app for Mac on November 19. Users describe a task in a single sentence, and Gemini clicks, types, navigates menus and fills in forms on its own, finishing the job across local files, desktop software and web pages. The first wave is limited to Google AI Plus, Pro and Ultra subscribers in the US using English.
The launch blog post is signed by Michael Friedman, product lead for the Gemini app, and it is short. The hardware requirements sit in a footnote: a Mac with Apple Silicon running macOS Sequoia 15.0 or later, and a personal Google account belonging to someone 18 or older. Work and school accounts are not covered this time. Users in Europe and the UK can't use it yet.
Working in a background window
The point Google stresses most in this launch is that Gemini does not fight the user for the mouse. It runs in an isolated background window, with a live preview in the chat box, while the user's own screen and cursor stay fully usable.
“Your main screen and cursor remain completely free, so you can keep typing, designing, or taking calls without interruption.”
You invoke it by adding an “@Computer use” tag in the conversation. Users choose in advance which apps Gemini may touch, can watch every step as it happens and can pause at any time. Key actions require the user's approval before they run. According to 9to5Google, Gemini also stops and waits for confirmation at sign-in and payment steps, and the settings now include desktop connectors for apps such as Blender and Autodesk Fusion.
Google's four examples are all everyday chores: batch-adjusting colors or building a 3D scene in an image editor; pulling weight, dimensions and addresses from local PDFs into a multi-column shipping quote form; capturing metrics from regional data dashboards and assembling them into a Google Slides deck; and analyzing photos of clothes in a folder, looking up prices, logging the details and listing the items on a resale platform.
The risk warning is equally blunt. Google says Gemini can make mistakes, including permanently deleting files or sharing them with other people. Backups of files that were changed are deleted after 24 hours, or when the user starts a new task.
From Spark to computer use
Strung together, Google's steps on the Mac this year come thick and fast:
- June 30: the Gemini Spark agent arrived on Mac, limited to US AI Ultra users aged 18 and over, and able to touch only folders the user authorized;
- Late July: Gemini on Mac gained voice control of apps that are already open, and Spark later expanded to AI Pro users in more than 160 countries;
- September 10: Gemini got a Windows client, summoned with Alt + Space;
- October 13: the Gemini app added a task mode that keeps running after the user walks away;
- November 17: task mode became free in the US and five other countries;
- November 19: computer use went live.
On the developer side, things came earlier. Google opened the Gemini 2.5 Computer Use model to developers in October 2025, and this June built computer-use capability into Gemini 3.5 Flash. What changes this time is the vehicle: the capability moves out of APIs and browsers into a consumer app on an ordinary person's computer, and the range of things it can operate widens from web pages to any desktop software the user names.
The pricing tiers reveal Google's plan. Task mode, which only handles files and connected services, is already free in the US, while computer use, which drives the screen directly, starts at the cheapest paid tier. Based on previously announced prices, AI Plus costs $4.99 a month, Pro $19.99, and Ultra more. PPC Land's rough count puts 142 days between June's Spark, offered only on the most expensive tier, and this computer-use launch.
What is still unclear
Google's announcement and its help documentation don't fully agree. AppleInsider pointed out that the support document published on launch day lists only Pro and Ultra as the plans that can control desktop apps, while the announcement includes AI Plus. Google hasn't said separately which one governs. TechCrunch reported that Google also hasn't said when it will expand to other countries or other operating systems.
Another gap concerns the web side. When Gemini opens a website or fills in a form from its background window, the announcement doesn't say whether the site sees an ordinary browser or an identifiable agent. For e-commerce and form-driven businesses, that affects risk controls and how traffic is counted.
Similar products are already plentiful. Anthropic's Cowork reached macOS in January as a research preview, and OpenAI's Chrome extension can operate browser tabs to fill in forms and send messages for the user. Gemini's difference this time is the “background window,” where the person and the agent each have their own cursor. How steady it is in daily use will depend on feedback from the first batch of Plus users.
Sources: Google Gemini official blog (feature scope, device and account requirements, four examples), TNW, CocoLoop, PPC Land (subscription prices and product timeline), 9to5Google (invocation and desktop connectors), AppleInsider, TechCrunch; Google Help Center (system version and age requirements verified).