Gemini gains Mac computer control, with rollout limited to U.S. users
Google says its Gemini app for macOS can perform tasks across files, apps and websites. Availability is initially limited by location, age, account type and subscription.

The report classified 60% of macOS users as power or mainstream AI users.6
The corresponding figure reported for iOS users was 52%.6
The corresponding figure reported for Windows users was 46%.6
Story
What Gemini’s Mac computer-use feature changesGoogle announced on Nov. 19 that its Gemini app for macOS can operate a Mac for a user, adding computer-use capability directly to the app.12 The feature, named Computer use, can work across files stored on the computer, desktop applications and websites.2 Rather than relying only on a user to navigate each screen, Gemini can respond to a task described in ordinary language by clicking, typing, using menus and completing forms.3 The announcement describes a new way to access computer-operating capability through the Mac Gemini app; it does not provide a numerical measure of how many users or tasks are involved.12
Gemini performs the work in a separate background window, while a live preview in the chat lets the user see what is happening.2 That separate window is intended to leave the user’s main display and cursor free for other activity during the process.2 The preview therefore provides visibility into the task while the user continues to have the main screen available, according to the described setup.2 For actions Google considers important, the system asks the user to approve them before proceeding.2
Google’s examples show how the feature combines activity on a Mac with work on the web: one task gathers figures and screenshots from regional dashboards and places them in Google Slides.1 This example involves collecting material from dashboards and assembling it in a presentation, rather than merely summarizing text in a chat.1 Another example starts with clothing photographs saved in a Mac folder, then has Gemini look up prices online and prepare listings for a resale service.1 That proposed workflow links local images to online research and preparation of sales listings, illustrating the feature’s reach across different kinds of digital work.12
Other suggested tasks include making changes to several photographs, creating elaborate three-dimensional scenes, and taking information from PDFs to complete shipping-estimate forms.2 These examples range from visual editing and scene creation to transferring information between a document and a form.2 Together with the dashboard and resale examples, they indicate the kinds of tasks Google presents, but they do not establish how often users will perform them.12 The feature’s described reach across local files, applications and websites is what allows those examples to combine steps that would otherwise occur in separate places.12
The initial rollout is in the United States and uses English, so the announced availability is geographically and linguistically limited.6 Google had not set out a schedule for offering the feature beyond the United States.7 Eligibility also depends on the user being at least 18, using a personal Google account and holding a Google AI Plus, Pro or Ultra subscription.1 These conditions narrow the initial audience by location, language, age, account status and paid plan; the announcement does not quantify how many people meet them.167
The feature follows earlier Google work on computer operation: the company had shown its Gemini 2.5 Computer Use model to developers in October 2025.2 Google later incorporated Computer use into Gemini 3.5 Flash in June 2026.2 Google had also made automated Mac file operations available through its Gemini Spark AI agent.2 The present change is that Computer use is accessible within the Mac Gemini app itself, alongside those earlier previews and agent-related offerings.2
Google is not the only company associated with software that can operate Mac applications and websites: a report named OpenAI, Anthropic and Meta among firms whose agents already did so.3 That report places Gemini’s addition in an existing field of computer-operating agents, rather than describing the capability as unique to Google.23 The comparison is about the capability to work with Mac apps and websites; the available facts do not supply a common measure of performance, adoption or market share.3 Accordingly, the presence of other providers offers context for the announcement, but does not establish how their products compare with Gemini in practice.3
Apple’s macOS 27.3 Beta 2 included revisions concerning permissions for device control and access to data, according to an Apple developer announcement dated Nov. 16.5 The report said those changes came before Google’s Gemini announcement and were not a short-term measure taken in response to this feature.5 The timing matters for interpreting the two developments: Apple’s permission revisions are related to Mac control and data access, but the report does not connect them as cause and response.5 The available account therefore places Google’s new app capability alongside Apple’s permission changes without presenting the latter as a reaction to Gemini.235
A PYMNTS Intelligence report put the share of macOS users classified as power or mainstream AI users at 60%.6 That reported proportion was higher than the corresponding figures for iOS users, at 52%, Windows users, at 46%, and Android users, at 40%.6 The figures compare reported user categories across four operating systems; they are not counts of Gemini subscribers or estimates of people eligible for Computer use.16 The report provides context for the Mac audience, while the rollout’s actual reach remains constrained by its U.S. launch and its account, age and subscription conditions.167
Bloomberg Technology characterized the announcement in its headline as Gemini gaining the ability to control Mac devices for users.8 The operational detail in the other reports is more specific: tasks may span files, applications and websites, with progress shown in chat and approval requested for important actions.2 Those details describe both the feature’s reach and parts of its interaction with the user, while the initial audience remains limited by geography and eligibility requirements.126 Taken together, the reports place the addition within a competitive field and a changing Mac permission environment, without claiming that either context determines how widely the feature will be used.35
Structure
Who is connected to whom- 1Google Gemini
- ← usesFact
- enables →Fact
- ← usesFact
- ← usesFact
- ← usesFact
History
How it came to this- October 2025Computer Use model previewed to developersGoogle previewed its Gemini 2.5 Computer Use model to developers.
- June 2026Computer use integrated into Gemini 3.5 FlashGoogle later integrated Computer use into Gemini 3.5 Flash.
- November 16Apple announces Mac permission changesApple's macOS 27.3 Beta 2 included changes to device-control and data-access permissions; the report said these were not a temporary response to Gemini.
- November 19Computer use added to the Mac Gemini appGoogle announced that its macOS app could operate a Mac on a user's behalf.
- NowAdded Gemini app for macOS gains Computer use
Impact
Spreading outward, level by level- Level 1User interaction
Users can describe tasks in natural language; Gemini can click, type, use menus and fill in forms. It works in an isolated background window with a live chat preview, leaves the main screen and cursor available, and asks for approval before important actions.23
Fact - Level 2Initial availability
The rollout is in the United States in English. Eligibility requires users to be at least 18, use a personal Google account and subscribe to Google AI Plus, Pro or Ultra; no timetable for availability outside the United States had been announced.167
Fact - Level 3Competitive context
A report named OpenAI, Anthropic and Meta among companies whose agents already supported operating Mac apps and websites. This places Google's addition in an existing field, without establishing comparative performance or adoption.3
Fact - Level 4Mac permission context
Apple's permission changes preceded Google's announcement and were reported not to be a temporary response to this feature; the available account does not establish a cause-and-response relationship.5
Fact
Sources
What each source supportsWritten by AI from the sources listed below: every fact was checked word for word against its source, and inference is marked apart. How we write