Gemini in the browser: where it fits next to your other AI tabs

A Gemini tab is rarely alone. It sits somewhere in a row of twenty or thirty, between a ChatGPT tab, a Claude tab, the Notion page being drafted, and the Slack thread that started all of it. The question is not whether Gemini is useful. The question is which of the several things called "Gemini in the browser" is actually running, what it can see, and where the window should live so that it stops disappearing behind the rest of the work.

Three doors carry the same name

Google ships Gemini into the browser through more than one route, and they are not interchangeable. The clearest statement of the difference comes from Google's own product page.

Gemini in Chrome is part of the Chrome browser on desktop. It opens from the Gemini icon in the Chrome toolbar, or from a keyboard shortcut set up by the user, on Windows, macOS and Chromebook Plus. Because it is part of the browser rather than a page inside it, it can work with the content of open tabs.

The Gemini web app is a site. It lives at gemini.google.com, and Google lists Chrome, Safari, Firefox, Opera and Edgium as the browsers that can reach it. Anything a normal page can do, it can do. Reading the tab next to it is not on that list.

The third door is a shortcut into the second. In Chrome, typing @gemini in the address bar and then pressing Tab or Space turns the address bar into a prompt box. Enter a prompt, press Enter, and the browser redirects to gemini.google.com with that prompt. Google's Chrome help page is explicit that this shortcut cannot be edited, deleted, or set as the default search engine, and that the shortcut can be switched off in Settings.

Google's product FAQ draws the line in one sentence: the web app can be used in other browsers, or in the content area of Chrome, but page content cannot be shared and Live mode is not available the way it is with Gemini in Chrome. That single sentence decides most setup questions.

What the browser-level version can see that a tab cannot

The whole argument for the in-browser version is context. Google describes it as delivering summaries of articles, pages and threads, turning a window full of open tabs into an organised comparison of specs and pros and cons, and answering questions based on what is currently open. On mobile the same idea arrives differently: Gemini can be activated on Android by holding the power button, and on iOS Google says it is being built into the Chrome app with access through the omnibox.

Two limits are worth reading carefully before rearranging a workflow around it.

The first is availability. The automated browsing feature, where Gemini carries out a multi step task such as booking an appointment on the user's behalf, is described by Google as rolling out first in preview to Google AI Pro and Ultra subscribers in the United States. Other suggestion features are described as rolling out to eligible users in select regions. The fine print on the same page reads: setup required, compatibility and availability vary, 18+.

The second is that it only runs when asked. Google states that Gemini in Chrome activates when the icon or the keyboard shortcut is used, not on its own, and that Gemini Apps Activity can be reviewed, deleted, or turned off at any time. For anyone whose browser holds client work, that control matters more than any feature on the list.

The practical reading is that the toolbar version is the only one with page context, and the page context is exactly what makes it worth a dedicated place on screen.

Comparing the entry points

Entry point	How it opens	Can it use the open page	Where it runs
Gemini in Chrome	Toolbar icon or a keyboard shortcut you set up	Yes, including Live mode	Inside Chrome on desktop
Gemini web app	Visit gemini.google.com	No	Chrome, Safari, Firefox, Opera or Edgium
@gemini in the address bar	Type @gemini, press Tab or Space	No, it redirects to the web app	Chrome, ending on gemini.google.com

The table is not a ranking. A team that works across Safari and Firefox has one sensible option, and it is the web app. A person who spends the day in Chrome and wants summaries of what is already open has a different one.

The account question that breaks most setups

Access rules are tied to the account, not the browser. Google's Chrome help page for the web app states two requirements: a personal Google Account managed by the user, which excludes an account managed by Family Link, and being 13 or over, or the applicable age in the country. Work and school accounts follow separate administrator rules, which Google documents on its own page for Gemini Apps on a work or school account.

That is where the trouble starts for anyone holding several accounts at once. Google's own documentation on signing in to multiple accounts says the settings are usually separate, but that when several accounts are signed in at the same time it cannot always tell which one is in use, and settings from the default account may apply. In day to day terms, the AI feature that works on the personal account behaves differently under the work account in the same window, and the account that is treated as default is the one that was signed in first.

Chrome's answer is profiles. A Chrome profile keeps bookmarks, history, passwords and settings apart, which is the recommended way to separate work from personal. The cost is that a profile is a whole browser window with its own set of tabs, so separating three accounts means running three browser windows and remembering which one holds which. A browser that keeps each web app in its own window with its own cookie container attacks the same problem from the other side, giving each account a window with a name instead of a colour swatch. The page on Workspaces describes that container model in more detail.

Where the AI window should actually live

An AI assistant is a tool that gets reached for dozens of times a day and is needed for a minute at a time. That usage pattern is the opposite of a document, and it is the worst possible fit for a tab in a long row. The tab that gets used most often is the tab that is hardest to find, because it sits among the ones that were opened and abandoned.

There are three ways to give it a fixed place, and they differ in how much they cost.

Install the page as an app

Chrome can turn a site into its own window. On a Mac, select More, then "Cast, save, and share", then "Install page as app". Installed web apps can be managed at chrome://apps, where right-clicking an app offers "Create shortcut" and "Launch at startup". This is the cheapest option and it works today. What it does not do is separate accounts: the installed app uses the profile it was installed from.

Give the window a keyboard shortcut

For the in-browser version, Google's own instruction is to set up a keyboard shortcut rather than hunting for the toolbar icon. The habit to build is reaching for a key, not for a tab.

Put every web app in a window of its own

The third option is a browser built around app windows instead of tabs, where each app keeps its own session and its own notification state. The Supported apps list shows the kinds of services people run this way, and the pattern that emerges is consistent: a handful of apps that are open all day, each with its own account, and none of them sharing a cookie jar.

When a second assistant is already open

Most people asking this question are not choosing an assistant. They already have two or three open, and they are trying to work out which one deserves the front row. The honest split is by what the tool needs in order to answer.

An assistant that reads the page needs the page. That argues for the in-browser version, and it argues for keeping it in the same window as the work rather than in a separate one, because the context it uses is whatever is open beside it.

An assistant that is being used as a writing surface, where a long thread is built up over a week and returned to, behaves like a document. Documents survive better in their own window with their own name in the Dock, because a thread that matters is a thread that should not be one accidental tab close away from being hard to find again.

An assistant that is being compared against another one needs both answers visible at once, which no single tab row does well. Two windows side by side, each holding one service, is the arrangement that makes a comparison readable.

Sorting tools by that question rather than by brand loyalty tends to settle the layout quickly. One gets the toolbar, one gets a window, and the rest can stay bookmarked.

The record of what was asked

Anything that reads the open page is reading whatever is on it, including a client's document or an invoice. Two controls are worth setting deliberately rather than discovering later.

Google states that Gemini Apps Activity can be accessed at any time to manage, delete, and turn off that activity. Turning it off changes what is retained for personalization, and it is a setting rather than a one time cleanup, so it is worth deciding once per account instead of per session.

The second control is the account itself. Because settings from a default account can apply when several accounts are signed in to one browser, the safest arrangement for anyone who works across a personal account and a client account is to keep them in separate containers, so that the question of which account was active never has to be reconstructed after the fact. Google's own multiple sign-in documentation admits the ambiguity, and admitting it is more useful than pretending a colour badge solves it.

For work covered by a client agreement, the order of operations is to check what the agreement allows, then set the activity control, then decide where the window goes. The layout question is the easy one.

What to change first

Decide which Gemini is actually needed. If page context matters, use the toolbar version in Chrome and set the keyboard shortcut in the same sitting, because the icon will be missed and the shortcut will not. If context does not matter, the web app in a dedicated window is steadier, and installing it as an app takes about a minute. If the reason the tab keeps getting lost is that several accounts are fighting over one window, the fix is separation rather than another tab, which is what SpaceDeck was built to do.