Chrome Auto Browse Explained: Gemini Tasks, Your Passwords, and Who Gets It

Google's personal AI agent can drive your actual web browser. On July 30, 2026, Google connected the agent then called Gemini Spark directly to Chrome, and the part worth stopping on is what it is allowed to use: your logged-in accounts and your saved passwords. Google renamed the agent to "Tasks in Gemini" in the week of October 13, 2026, so its help pages now say Gemini rather than Spark; nothing about what auto browse does changed with the name. Our Tasks in Gemini guide covers the rename in full. Google's example errands are scheduling apartment viewings and researching flights and starting the booking. This page explains what Chrome auto browse actually does, what you need to have it, where it works, and the controls you should know about before you switch it on.

Key Takeaways

• Chrome auto browse lets Gemini use your own browser, signed in as you. Per Google's tasks help page, re-read October 16, 2026, it can reach "all the same sites that you do, including sites you're signed into," and with your permission use "login info you saved in Password Manager to sign into loyalty programs and online accounts."
• You need Google AI Pro or Google AI Ultra. There is no free tier for this, and Google AI Plus does not include tasks. Since mid-October 2026 you must also opt in to Personal Intelligence -- a new line in Google's requirements list, which means Gemini reads your Gmail, Contacts, Photos, Search and YouTube history before it will take a task at all.
• The task version is desktop Chrome only, and Google's wording changed with the rename. It used to read "only Gemini Spark in desktop Chrome can use auto browse." As re-read on October 16, 2026 it reads "Currently, Chrome auto browse is only available on desktop," and a new sentence says Gemini "only uses the remote browser on mobile devices." Chrome's own agent is a different surface and reached Android on August 18, 2026 -- so "auto browse is desktop only" is true of Gemini driving your browser, not of the feature as a whole.
• Your paid plan caps how much of it you get. Google's Chrome help page, re-read September 11, 2026, publishes a daily ceiling on multi-step tasks in Gemini in Chrome: "With a Google AI Pro plan: up to 20 requests per day. With a Google AI Ultra plan: up to 200 requests per day." Tasks are capped differently -- the tasks help page, re-read October 16, 2026, still describes "compute-based usage limits" with no published count, and still allows up to 15 tasks running at once ("a schedule will not run if there are already 15 tasks running").
• The US on a computer, the US or India on Android. Google's July 30 announcement said the Chrome capabilities "are initially rolling out in the U.S. with plans to expand to additional regions in the future," and the first expansion has happened on one platform: its auto browse page, read October 7, 2026, requires you to be "18 or over and in the US" on a computer but "18 or over and in the US or India" on Android.
• The agent itself went nearly worldwide on the same day -- Google AI Pro subscribers in "over 160 additional countries" -- but it remains blocked in the European Economic Area, the UK, Switzerland and Nigeria, in that exact wording on an October 16, 2026 re-read. Full picture in our Tasks in Gemini guide.
• You approve it twice, then per task. Chrome asks permission the first time on a device, and Gemini asks for confirmation before every browsing task after that.
• Payments get handed back to you. Google's July 30 announcement says the work is done "while protecting against threats like prompt injection and keeping you in the loop on sensitive actions, such as payments, by handing the task back to you."
• The honest verdict: the most capable consumer AI agent step so far, and the one with the most to lose if it goes wrong. Worth trying on low-stakes errands if you already pay for Pro or Ultra. Do not point it at anything involving your money.

What Is Chrome Auto Browse, in Plain Terms?

It is Gemini using a real web browser the way you would, rather than reading the web from a distance. Google describes the jobs as "tasks that involve browsing the web, like shopping, booking travel, or making dinner reservations." Because it can work inside your own signed-in Chrome, it is not stuck outside the login wall the way a normal AI chatbot is -- it can get to the pages that require an account.

The everyday version: you assign a task -- "find me three two-bedroom flats near the station under this rent and book viewings" -- and instead of handing you links, it opens the sites, fills the forms, and comes back when it needs you. That is the leap. It is also exactly why the permission model below matters more than the feature list.

Which Browser Does It Actually Use? Local vs Remote

Two different things can happen, and the difference decides what Gemini can see. Google's help page splits them explicitly:

Google's wording on the local browser is worth reading twice, and as of the rename the subject of the sentence is Gemini rather than Spark: "When acting on your behalf, Gemini will share necessary info with third parties. This could include your name, contact information, files, preferences, and info you might find sensitive." That is the trade. The remote browser is the more private option and the less capable one, because it hits a wall at every sign-in page.

Do You Get It? Availability by Plan and Place

The requirements are the same as Tasks in Gemini's, plus a US location and a desktop Chrome. Here is who gets what, per Google's tasks help page and the July 30 announcement, re-read October 16, 2026:

Four more conditions apply whatever you pay: you must be 18 or over, signed in with a personal Google Account, have Keep Activity on, and -- new since mid-October 2026 -- have opted in to Personal Intelligence. For Chrome auto browse there is a fourth that is easy to miss, and a reader who has turned it off will find the feature simply absent: Google requires you to set Safe Browsing to Enhanced or Standard protection in Chrome, on both the computer and the Android variants of its auto browse page, read October 7, 2026. It also does not run in Incognito mode, and on a work or school account your administrator has to enable it. And four places are excluded from tasks outright, so auto browse cannot reach them either: the European Economic Area, the United Kingdom, Switzerland and Nigeria. A UK reader on Ultra still has nothing to switch on. If you are in the UK and want to know what Gemini in Chrome does give you there, that is a different and much smaller feature -- see Gemini in Chrome in the UK.

One note on where Google says what. gemini.google/subscriptions, read October 16, 2026, sells Ultra on "first access to advanced features like Tasks in Gemini (in select countries)," which reads as if Ultra is the only way in; the phrase "Deep Think" no longer appears anywhere on that page, where it used to sit in the same sentence. The help page and the July 30 announcement are newer and more specific, and both say Google AI Pro qualifies. Google One's plan page agrees, putting "Access to Tasks in Gemini" on the Pro card. Follow those.

How Much Does It Ask Before Acting?

More than you would expect, and Google has published the exact points. There are three separate gates:

You can also watch and stop it mid-run. Google moved where you look for this with the rename: it used to be a Recent Tasks list, and the October 16, 2026 wording is that you "open the task thread to check if Gemini is using a browser for that task," with a View live link next to "Gemini is using Chrome". There is still a Stop button in the Gemini in Chrome side panel. Google warns that stopping does not necessarily end the job: Gemini "tries to figure out the next steps to continue your task," which may mean switching to the remote browser, using a different tool, or abandoning the task. To cancel outright, use the Stop button in the Gemini text box instead.

To switch the whole thing off on a device: More > Settings > AI Innovations > Gemini in Chrome.

Is It Safe to Let It Use Your Passwords?

Only with limits, and Google names the biggest one itself: do not type sensitive details into the chat. Its help page says plainly, "Do not enter sign-in info, payment details, or info you may find sensitive directly into the task thread." If a login is genuinely needed, Google's instruction is to take control of the browser and type it on the website yourself.

Google also says the Chrome integration is built "while protecting against threats like prompt injection" -- the attack where a web page hides instructions that hijack an AI agent reading it. That is the right risk to be defending against, and no vendor has claimed to have solved it. The practical rules from our AI agent safety basics apply here more than anywhere: keep it away from banking and your primary email, approve every purchase and outgoing message yourself, and check what it did rather than trusting the summary. Google's own guidance says the same in weaker words -- "avoid tasks or schedules that use information or perform actions that you find sensitive."

There is one scenario worth planning for. If a scheduled task runs while you are offline, Google warns "you may not be able to stop Gemini from completing an unintended action." So a standing overnight instruction is a different risk from a task you sit and watch.

Should You Turn It On?

If you already pay for Google AI Pro or Ultra, you are in the US, and you have a genuinely tedious errand -- comparing flight options, filling out the same viewing-request form on nine listings -- yes, try it on that. It costs nothing extra, the approval gates are real, and this is the first consumer agent that can actually get past a login wall.

Skip it if the task involves money leaving your account, your primary email, health or financial records, or anything you would not hand to a competent stranger on their first day. And do not upgrade to Pro or Ultra for this -- the Chrome half is US-only, and the wider agent -- Tasks in Gemini -- is still a supervision-hungry tool rather than a hands-off one. Our guides to ChatGPT vs Gemini for everyday use and which AI chatbot is worth paying for cover whether the plan is worth it on everything else it includes.

The bottom line: Chrome auto browse is the most capable consumer AI agent feature shipped so far, and Google has been unusually specific about its permissions, its fallbacks and its off switch. Use it for errands, not for money. And if you are outside the US, this one is still a preview of something you cannot have yet.

Corrections and Updates

October 16, 2026: Google renamed the agent this page is about. "Gemini Spark" is now "Tasks in Gemini", and a new requirement was added: you must opt in to Personal Intelligence before you can assign a task. The help article was re-read live on October 16, 2026; the Internet Archive copy of October 9, 2026 was still titled "Use Gemini Spark". Nothing about auto browse's own behaviour, plan gate or geography changed -- Google AI Pro or Ultra, 18 or over, personal account, Keep Activity on, desktop Chrome, US on a computer and US or India on Android, Safe Browsing Enhanced or Standard, no Incognito, and the EEA, UK, Switzerland and Nigeria still excluded, all re-read and held. Three wordings did move and the page now carries the current ones: "only Gemini Spark in desktop Chrome can use auto browse" became "Currently, Chrome auto browse is only available on desktop", with a new sentence that Gemini "only uses the remote browser on mobile devices"; the local-browser data-sharing quote now has Gemini rather than Spark as its subject; and watching a running task moved from a "Recent Tasks" list to a "View live" link in the task thread. The 15-concurrent-task cap and the 20-and-200 per-day Chrome figures both re-confirmed verbatim. The title, summary, lede, all six FAQ answers, the Key Takeaways and the approval table were updated; quotations from pre-rename documents keep their original words.

October 7, 2026: this page said Chrome auto browse is "US only for now" on every surface, and it omitted the Safe Browsing requirement. Google's auto browse page, read October 7, 2026, says "Be 18 or over and in the US" on the computer variant but "Be 18 or over and in the US or India" on the Android one, and both variants require you to "set "Safe Browsing" to Enhanced or Standard protection" -- a prerequisite this page never carried, and the one most likely to make the feature look missing. The plan table, the Android FAQ and the conditions paragraph now say so, and the work-or-school cell is corrected from a flat No to "only if your administrator enables it", which is what Google actually publishes. Spark's own gates all held on the same read: Google AI Pro or Ultra, 18 or over, personal account, Keep Activity on, desktop Chrome only, and the EEA, UK, Switzerland and Nigeria exclusions.

September 11, 2026: re-verified, no claim changed. Every load-bearing claim was re-read against its primary source on September 11, 2026 and held: the Password Manager permission, the "only Gemini Spark in desktop Chrome can use auto browse" scope, the Google AI Pro or Ultra requirement, the 18+ / personal-account / Keep Activity conditions, the EEA, UK, Switzerland and Nigeria exclusions and the off switch, all against the Spark help page; and the prompt-injection, payments and "initially rolling out in the U.S." quotes against the July 30 announcement. The 20-and-200-per-day figures were re-confirmed verbatim on the Chrome help page, which the August 18 note below had never cited; that citation is now in the key takeaway. Two facts were added from the same read: Spark allows up to 15 tasks running at once, and Gemini may pause and ask you to take control of the browser to type a password or payment detail yourself.

August 18, 2026: "auto browse is desktop only" needed scoping, and there are now published daily limits.