Gemini Mac application: which one you are looking for
The phrase carries at least three destinations. One is Google's AI assistant, shipped as a macOS build with a narrow hardware requirement. One is a duplicate file cleaner that has carried the same name for a decade and has nothing to do with AI. The third is not a product at all: it is the option of taking gemini.google.com and giving it the shell of an application, which is what most people are really after when they type this. Sorting out which one is on screen takes about a minute, and the minute is worth spending, because the three have almost nothing in common.
Three products answer to the same name
The fastest way to tell them apart is the machine each one runs on.
| Product | What it is | Runs on | Cost |
|---|---|---|---|
| Gemini for macOS (Google) | AI assistant with hotkeys and screen context | Apple silicon MacBooks, macOS Sequoia 15.0 or later | Free |
| Gemini 2 (MacPaw) | Duplicate and similar file finder | macOS 10.10 and higher | From $1.95 per month |
| A self-built shell around gemini.google.com | A window that only opens one site | Any Mac running a Chromium browser | Free for up to three apps |
MacPaw's product is the older resident of the name, and its listing is unambiguous about what it is for.
System Requirements: macOS 10.10 and higher, 115 MB. Pricing: Starting at $1.95/month. Latest Version: 2.10.1, 18 June 2026. Source: macpaw.com
Note the gap in the first column. A file cleaner released years ago still supports macOS 10.10. Google's AI assistant, released far more recently, requires macOS 15 and an Apple silicon MacBook. Age has nothing to do with it. Newer software from large vendors tends to set a higher floor, not a lower one, and that floor is the single fact that decides this search for a large share of the people typing it.
What macOS grants an application that it never grants a tab
Before comparing features, it helps to be precise about what the word application buys, because the list is mechanical rather than aesthetic.
An application is a bundle in the Applications folder with its own identity. That identity produces a set of behaviours that a browser tab cannot reach. It occupies a slot in the Command and Tab switcher, so it can be reached without touching the mouse. It holds a permanent position in the Dock, in a place the eye learns. It appears in Spotlight and in Launchpad under a name the reader chose. It gets its own row in System Settings under Notifications, which means its alerts can be shaped or silenced without touching every other website. It appears separately in Force Quit and in Activity Monitor, so a stuck session can be ended without taking the browser down with it. It can be added to Login Items and be running before the first coffee.
A tab has an identity only inside the browser. Outside it, the tab is not addressable. Quitting the browser closes it, Command and Tab cannot reach it, the notification settings for it are buried under a site permission list, and its position in the tab strip moves every time something else is opened. None of that is a criticism of browsers. It is a description of where the boundary sits.
This is why the same reader can be well served by two very different answers. Someone who wants Gemini to read the window in front of them needs system permissions, which only a native application can hold. Someone who simply wants Gemini to stop being the fourteenth tab needs an addressable window, which is a much smaller thing to arrange.
What Google's application asks for, and what it returns
The official build is free in every language and country where the Gemini app is supported. It runs only on Apple silicon MacBooks and requires macOS Sequoia 15.0 or later, so Intel machines and anything held back on macOS 13 or 14 are outside its range.
In return it offers three things a page cannot do. Option and Space opens a compact chat over whatever is on screen, and Option, Shift and Space opens the full window, with both shortcuts editable in settings. A menu bar icon and a Dock icon give two more entry points. Pressing both Command keys hands the contents of the frontmost window to the chat as context.
That third capability is where the trade sits. Google's documentation states that reading a full page in a browser, or reaching every file inside a shared local folder, requires enabling Accessibility for Gemini in System Settings under Privacy and Security. Accessibility on macOS is a broad grant rather than a per feature switch. It deserves a deliberate decision rather than a reflexive click, particularly on a machine that has client work on screen during the day.
Two of the more publicised features are also staged. Dictation into other applications, triggered by holding the FN key, is rolling out in English first. Gemini Spark, which organises connected folders and summarises their contents into Docs and Sheets, is being released first in English to Google AI Ultra users aged 18 and over in supported countries. For anyone working in another language, those two belong on a waiting list rather than in a decision.
What does not change between routes is the conversation itself. Google confirms that history and memory sync across desktop, web and mobile as long as the same Google account is signed in. The application is not a separate product with separate data. It is a different door into the same account, which is why the choice can be made on placement and permissions rather than on fear of losing anything.
Running one service as an application and a tab at the same time
This is the failure mode that follows a fresh install, and it is easy to walk into.
Once the application exists, the old tab usually survives. Now the same service has two homes with two separate cookie stores. Signing out in one leaves the other signed in. Notification permission granted in one place does not carry to the other, so alerts either arrive twice or arrive nowhere. A link to a conversation sent by a colleague opens in the browser copy, because the operating system hands http links to the default browser, not to the application that happens to display the same domain.
Three habits keep it clean. Decide which copy is the home for that service and close the other. If both are genuinely needed, give them different accounts rather than the same one, which turns a duplicate into a division of labour. And check System Settings under Notifications after the change, because the leftover permission is what produces double alerts weeks later when nobody remembers granting it.
When the official application cannot be installed
For an Intel Mac, or an Apple silicon Mac deliberately parked on an older macOS, the website is the only route to Gemini itself. The remaining question is not which AI to use. It is where the window lives.
A site to app tool addresses exactly that and nothing more. The mechanism is worth understanding before choosing one. Rather than bundling a browser engine of its own, this kind of tool builds an application bundle that points at a Chromium browser already installed on the Mac, using a symlink to the real binary. Chrome, Brave, Edge, Opera, Vivaldi, Chromium and Arc can all serve as the base, and because the reference is a link rather than a copy, updating the browser updates the window. Each generated app carries its own profile, so its cookies and extensions are separate from the browser and from every other generated app. Notifications, Dock presence and the tab bar are toggles rather than assumptions. The catalogue of ready made entries is worth a look before building anything by hand, since more than three hundred services are already described with their URL, name and icon, and the full list sits on the Supported services page while the option by option reference is in the Guide.
The honest limits matter as much as the capabilities. A window built this way does not claim a system wide hotkey, cannot read the frontmost window of another application, and has no access to local folders. Those abilities come from permissions granted to a native application, not from the shape of a window. What is gained is narrower and more durable: a fixed place in the Dock, a real entry in the app switcher, a session that survives quitting the browser, and an icon that is recognisable at a glance.
A test for which tabs deserve an application
Not every site earns one. Five questions settle most cases, and a site that answers yes to three or more is a candidate.
Is it opened every working day, rather than looked up occasionally? Does it need to stay signed in as a specific account while another account is signed in elsewhere? Does losing it hurt when the browser is quit or crashes? Does it send notifications worth receiving at the operating system level? Is it something reached by reflex, where hunting through a tab strip is a measurable interruption?
Gemini scores well on the first and second for heavy users, and poorly on the third and fourth. A shared inbox, a project tracker or an admin console usually scores higher. That is the useful output of this exercise: the reader arrived asking about one service and leaves with a short list, which is a better answer than a single yes. Working through the list takes a few minutes, and the Features page shows which of the behaviours above are toggles rather than fixed choices.
A worked example makes the second question concrete. Consider someone holding a personal Google account and a company account with Gemini enabled by an administrator. Signing in and out all day is the usual answer, and it is the expensive one, because every switch costs a redirect, a password manager prompt and a lost train of thought. Two windows, each holding one account permanently, removes the switch entirely. The same logic applies to a freelancer holding an account for each client. The number of windows is set by the number of identities, not by the number of services.
The third question deserves a note as well. Browsers are restarted more often than people think, by updates, by a crash, or by the daily habit of closing forty tabs at once. Anything that has to survive that event is better placed outside the browser window, and anything that does not is fine where it is.
What to change first
Check the chip and the macOS version in About This Mac before anything else, because that single line decides whether Google's application is even available. If it is, install it and leave the old tab closed. If it is not, pick the one or two sites that fail the tab test most badly and give them their own windows with Kagemusha, which is free for the first three.
Frequently asked questions
Is the Gemini Mac application from Google free?
Yes. Google describes it as available at no cost in all languages and countries where the Gemini app is supported. Some features layered on top, such as Gemini Spark, are tied to a paid Google AI subscription and are rolling out gradually, but the application itself carries no price.
Why will the Gemini application not install on this Mac?
The most common reason is hardware. The build runs only on Apple silicon MacBooks and requires macOS Sequoia 15.0 or later, so every Intel Mac is excluded regardless of memory or purchase date, as is any Apple silicon Mac still on macOS 13 or 14. The second most common reason is an account one: a work or school account needs an administrator to have enabled Gemini.
Is the Gemini duplicate file finder related to Google Gemini?
No. Gemini 2 is a duplicate and similar file finder from MacPaw, sold separately and running on macOS 10.10 and higher. The two products share a name and nothing else, which is why search results for this term mix AI articles with disk cleanup articles.
Can a browser tab be turned into something that behaves like an application?
Partly. A dedicated window can take a Dock position, an app switcher slot, its own profile and its own notification settings, which covers most of what people mean when they ask for an app. It cannot take system level abilities such as a global hotkey or reading another application's window, because those are granted to native applications rather than to windows.