Tasks in Gemini Is Now Free: Setup, Permissions, Purchases and Safety Limits

Google is making Tasks in Gemini available without a Google AI subscription for eligible adults in the United States, Canada, Australia, India, Japan and South Korea. Announced November 17, 2026, the rollout moves Gemini beyond answering a prompt: a task can continue in the background, use connected services, interact with websites and notify the user when the result is ready.

Free does not mean unrestricted or risk-free. Google requires a personal account with History on, limits some purchase and payment features to the United States, and warns that tasks can make mistakes or act while the user is offline. The useful question is therefore not just what Gemini can do, but which permissions, browser session, schedule and confirmation rules should be in place before it starts.

Tasks in Gemini at a glance

  • Launch: the free rollout began November 17, 2026, using Google’s Gemini 4 Hydrogen model.
  • Free regions: Australia, Canada, India, Japan, South Korea and the United States for eligible personal accounts.
  • Age and account: the user must be 18 or older, use a personal Google Account and keep Gemini Apps History on.
  • Where it runs: Google currently lists the Gemini mobile app, Gemini app on Mac and gemini.google.com for general tasks; purchase documentation additionally lists the Windows app.
  • What it can use: Connected Apps, skills, schedules, chats, logged-in websites, Personal Intelligence, location, a remote browser, a local Chrome session and a remote computer.
  • Purchases: eligible U.S. users can define merchant, item, budget and deadline conditions, with purchase confirmation on by default.
  • Control: the user can deny a permission, allow it once, allow it for the thread or save an always-allow preference where offered.
  • Main caveat: tasks can run when the user is offline, and turning task mode back on can automatically resume paused tasks and schedules.

Tasks in Gemini is not the same as an ordinary chat

Tasks in Gemini is not the same as an ordinary chat

Task mode can work across services, but the user should approve only the permissions needed for the current job. Illustration: ReaderSpice.

A chat normally waits for the next message and returns an answer in the conversation. A task can keep working, open a browser, gather information, create files, change content in a connected app or follow a schedule. Google gives examples such as cleaning up newsletters, preparing a post-vacation to-do list, watching flight and hotel prices, creating an itinerary and packaging research into a document, spreadsheet, presentation or PDF.

The distinction matters because a task may carry authority from the account and browser session that started it. If Gmail, Calendar, Drive, Photos, Wallet or a third-party service is connected, the agent can encounter personal information and actions that a normal web search would never see. Treat a task as a temporary assistant with scoped access, not as a smarter search box.

How to start a task safely

  • Open the Gemini mobile app, Gemini app on Mac or gemini.google.com with the personal Google Account you intend to use.
  • Confirm that Gemini Apps History is on and that the Task control appears beside the normal Chat mode. A staged rollout can delay visibility even for an eligible account.
  • Choose Task, then describe one outcome, a deadline and clear boundaries. Name the services Gemini may use and those it must avoid.
  • Start with a reversible task such as comparing travel options, organizing a copied folder or drafting a document. Do not begin with deletion, money movement or an important external message.
  • Review each browser or Connected App permission. Prefer Allow once until the behavior is predictable; do not select Always allow merely to clear a prompt.
  • If the task needs a sign-in, payment detail or other sensitive field, take control of the browser and enter it on the website. Do not paste the secret into the task thread.
  • Keep the first run supervised. Use pause, stop or take-control options when the plan changes, the wrong account appears or a site asks for more information than expected.
  • Inspect the final result and any external changes before treating the task as complete. A completion notification is not proof that every detail is correct.

Permissions: allow once before always allow

Google says Gemini can ask for permission before it uses a browser, makes a change in an app, performs an action or shares information. The choices vary by service. Do not allow declines the step; Allow once approves only that action; Allow for this chat can cover the rest of the current thread for connected Google products; Always allow updates the saved preference for future task threads.

The safest default is the narrowest temporary grant. Calendar read access may be reasonable for finding an open time, while permission to send mail or delete Drive files changes the consequence of a mistake. A single task that needs to read an itinerary does not automatically need access to every email, photo, contact, payment card and document associated with the account.

Review saved permissions after testing. Deleting a task thread can reset permissions granted only for that thread, while an Always allow choice remains in Connected Apps or browser settings until changed. If the prompt unexpectedly expands from reading to sending, deleting or paying, stop and create a new task with clearer boundaries instead of approving the escalation in place.

Remote browser, local Chrome and remote computer are different

Gemini can use a separate remote browser in the cloud or connect to a local Chrome session on a desktop. A remote browser can retain website authentication information and page content for later tasks. A local browser can use sites already signed in on the user’s machine, but Google notes that the device and Chrome may need to remain awake and running for the task to continue.

A remote computer is broader than browsing. Google’s privacy documentation says it can save and execute code, Markdown files and other task data. That saved information can include material from Personal Intelligence, Connected Apps or uploaded files. Do not give a remote computer a sensitive dataset simply because the final output is a harmless-looking summary.

The surface matters when troubleshooting. A task that works in the web app can fail on a sleeping local Chrome session. A remote browser may ask for a new sign-in after Google deletes saved browser data for privacy. Before repeating the task, check which browser or computer Gemini selected and whether the required session is still available.

Schedules can act while you are offline

Tasks can include schedules for recurring work such as a morning brief, price check or weekly research package. Google distinguishes these from scheduled prompts in ordinary chats. A schedule is an automated trigger for a task, so it can combine timing with browser, app and file actions instead of merely posting a reminder.

That power raises the supervision risk. Google warns that a scheduled task may run while the user is offline and may complete an unintended action before it can be stopped. Use recurring automation only for low-impact, reviewable work. Require confirmation for messages, purchases, bookings, deletion, sharing, publication and account changes, and periodically remove schedules that are no longer useful.

Turning task mode off deletes remote browser and remote computer data and pauses tasks and schedules, but it does not delete the schedules or task chats. Google says turning task mode back on automatically resumes the paused tasks and schedules. Review the Scheduled page before re-enabling task mode so an old automation does not restart unexpectedly.

Purchases and payments: keep confirmation on

Purchases and payments: keep confirmation on

Purchase confirmation is on by default and is the safer setting for a new agentic shopping workflow. Illustration: ReaderSpice.

Google’s purchase feature is narrower than the general free rollout. It currently requires an adult in the United States, a personal Google Account, History on and a payment method saved to Google Wallet. Google lists English support and access through the Gemini mobile app, Mac app, Windows app and web app. Work and school accounts are not supported.

The setup can include exact product details, permitted merchants, a maximum budget of up to $1,000, a shipping address and an expiration date within 14 days. Purchase confirmation is on by default. When it stays on, Gemini notifies the user when an order is ready and the user reviews it before paying with Google Pay. Turning confirmation off lets Gemini complete the transaction automatically when its conditions are met.

A safer purchase checklist

  • Keep purchase confirmation on for every new merchant, product type or booking workflow.
  • Use an exact model, color, size, quantity, seller preference and all-in budget rather than a broad request to find the best option.
  • Review shipping, cancellation, refund and warranty terms on the merchant page, not only Gemini’s summary.
  • Check the final total for taxes, fees, add-ons and recurring subscriptions before approving payment.
  • Cancel a task immediately if the merchant, dates, route or product changes. Google says an existing purchase task must be cancelled and recreated when its conditions need editing.
  • Review receipts in Google Wallet and keep the merchant’s own receipt; the Wallet record does not replace the seller’s official documentation.

Google says a transaction may use a secure virtual card number and may carry an extra bank-statement descriptor. Those protections reduce exposure of the underlying card number, but they do not verify that the product, merchant, fare rule or delivery promise is correct. Confirmation remains the decision point.

Data, privacy and deletion controls

Tasks can process task instructions, schedules, skills, remote-browser sessions, remote-computer files and data from Connected Apps, connected folders, Personal Intelligence and logged-in websites. Google says completing a task can require sharing necessary information with other services and third parties, including a name, address or other sensitive details. The requested action determines the data path; the Gemini interface is not a private boundary around every site it visits.

Remote-browser data can include cookies and page content. Deleting that saved browser data signs the user out of sites in the remote browser. Remote-computer data and code-execution history can be cleared separately in Task settings. Task activity can also be managed through Gemini Apps Activity. These controls should be part of cleanup after a sensitive experiment, not an afterthought when an error occurs.

Avoid passwords, authentication codes, full card details, health records, legal documents, confidential work files and private messages in a task thread. If a website needs a secret, take control and type it into the site. If the task does not need the data at all, disconnect the app or use a clean account rather than trusting the prompt to ignore what it can access.

Good first tasks and tasks to avoid

Good first tasks

  • Compare public flight or hotel options and produce a reviewable table without booking.
  • Summarize newsletters into a draft cleanup list without archiving or unsubscribing automatically.
  • Create an itinerary from public information and a copied, non-sensitive preference list.
  • Research a product category and produce a comparison document with source links.
  • Draft calendar changes or messages for approval instead of sending them.

Avoid or require direct supervision

  • Purchases, bill payments, reservations or travel changes without confirmation.
  • Deleting email, Drive files, photos, contacts or account data in bulk.
  • Sending messages, publishing posts or sharing documents to external recipients.
  • Handling medical, legal, tax, employment, immigration or financial decisions.
  • Using a logged-in password manager, authenticator, banking site or administrator console.
  • Running downloaded code or shell commands from an untrusted webpage or document.

Why Tasks in Gemini may be missing or fail

  • The account is a work or school Google Account instead of a personal account.
  • The user is under 18, Gemini Apps History is off or the rollout has not reached that account.
  • The country or plan is outside the current free or paid availability list.
  • The selected client does not support the requested task; Google’s general and purchase help pages list different app coverage.
  • The local Chrome browser or device went to sleep, closed or lost the signed-in session.
  • A required Connected App is disabled or a saved permission was denied or removed.
  • The remote browser data was cleared and the website now requires a fresh sign-in.
  • A merchant blocks automation, changes its checkout flow or requires an unsupported verification step.
  • A purchase condition changed; Google requires cancelling and recreating the task instead of editing it in place.
  • The task hit a safety restriction, confirmation gate, schedule expiration or product-specific limit.

A relevant Pixel 10 Pro example on Amazon

Tasks in Gemini does not require a new phone, and eligible users can start from the web or another supported client. Readers already shopping for a current Android flagship can compare the Google Pixel 10 Pro 128GB unlocked in Obsidian on Amazon (ASIN B0FFTT2J6N). The independently indexed Amazon listing identifies that ASIN as the unlocked 128GB Obsidian model. Google’s official specifications list a 6.3-inch display, Gemini apps and seven years of operating-system, security and Pixel Drop updates. Confirm the exact seller, storage, color, warranty, return policy and current price before ordering.

A Pixel purchase does not bypass account, age, region, History, permission or feature-rollout requirements for Tasks in Gemini. The safest reason to buy the phone is the full hardware and software package—not an assumption that one staged AI feature will always be available.

Affiliate disclosure: ReaderSpice may earn a commission from qualifying Amazon purchases made through this link, at no additional cost to you.

The bottom line

Making Tasks in Gemini free is a meaningful expansion because background work, schedules, connected apps, browser control and file creation are moving from paid access to a much larger audience. Used carefully, the feature can save time on research, organization and monitoring without requiring the user to supervise every intermediate step.

The risk also scales with access. A task can see more than a chat, continue after the screen is closed and cross from recommendation into action. Start with reversible work, use allow-once permissions, keep purchase confirmation on, supervise the first run, review schedules before re-enabling task mode and reserve sensitive or consequential decisions for a person.

Sources

Comments

comments