The short answer
Tasks in Gemini is a mode in Google's Gemini app that hands Gemini a multi-step goal, which it works through across Gmail, Drive, Calendar and the web while you are away. Since 13 October 2026 you switch to it with the Task toggle by the prompt box. It needs a personal Google Account, age 18 or over, Google AI Pro or Ultra, Keep Activity on, and is not available in the UK, EEA, Switzerland or Nigeria.
| Made by | |
|---|---|
| What it is | A Task mode in the Gemini app for multi-step work that runs while you are away (renamed from the Gemini Spark beta) |
| Toggle launched | 13 October 2026 |
| Who gets it | Personal Google Accounts, 18 or over, on Google AI Pro or Ultra; work and school accounts not yet |
| Not available in | European Economic Area, Nigeria, Switzerland, United Kingdom |
| Where | Gemini web app, Mac app and mobile app |
| Must be on | Keep Activity and Personal Intelligence |
| Running at once | Up to 15 tasks |
| Different from | Schedules in chat (up to 10 active) and the Gemini agent for work announced by Google Cloud |
What Tasks in Gemini is, and what it replaced
Tasks is the part of Google's Gemini app that does work for you instead of answering a question. You switch the prompt box from chat to task, describe a goal, and Gemini plans the steps, uses your Google apps, browses the web and creates files, then keeps going while you do something else. Google introduced the toggle on 13 October 2026, in the same announcement that brought Gemini 4 Argon to Google AI Pro and Ultra subscribers, and described it as Gemini working for you "even when you're not at your computer".
If you tried Gemini Spark earlier this year, this is the same feature with a new name. Google's help page says Spark was a beta that has been renamed Task, and that your existing tasks and schedules were kept.
Google's own examples show the kind of job it is built for: decluttering your inbox by summarising or archiving newsletters and unsubscribing from mailing lists, a custom news digest that follows the stories you care about, flight and hotel price alerts within a budget, a to-do list after a holiday, and research packaged into files you can download. The common thread is work with several steps that you would otherwise do by hand across Gmail, Drive, Calendar and a browser.
Three Google features sound alike, so it helps to separate them now:
- Tasks (this guide): multi-step work in the Gemini app for personal accounts on Google AI Pro or Ultra. A task can run once, at a set time, or when an event happens.
- Schedules in chat (Google calls them scheduled actions or Schedules): a recurring prompt, such as a daily news digest, that Gemini answers at a time you choose. You can have up to 10 active schedules, they need Keep Activity on, and they also work for qualifying Google Workspace accounts. Google's help page states plainly that schedules in tasks are different from schedules in chats.
- The Gemini agent: a separate "universal agent for work" that Google Cloud announced on 8 October 2026 at its Gemini at Work event, alongside Gemini Enterprise. Google gave no availability date. It is not the Tasks toggle in the Gemini app.
What you need before you start
- A personal Google Account and an age of 18 or over. Work and school accounts are not supported yet, so a Google Workspace login will not show the Task option.
- Google AI Pro or Google AI Ultra. Google's help page lists either plan. Note that the Google AI plans page lists earlier access to Tasks as an Ultra benefit in select countries, so on Pro the rollout may reach you later. The free plan and Google AI Plus do not include it.
- A supported country. Tasks is not available in the European Economic Area, Nigeria, Switzerland or the United Kingdom. Google does not publish a list of supported languages for Tasks.
- Keep Activity on, and Personal Intelligence switched on. Both are requirements. Keep Activity on a personal account means Google can use your activity to improve its services, including training models, and human reviewers may read some chats. Read our guide to keeping your data private in AI tools before you hand Gemini anything confidential.
- The right app. Tasks works in the Gemini mobile app, the Gemini app for Mac and the web app at gemini.google.com. The Windows app is not on Google's list.
- Chrome on a computer, if the task needs your own logged-in sites. Gemini can browse in a remote browser that keeps running after you close your device, or in your own Chrome on a desktop, where it can use every site you are signed into. For the local option, your computer and Chrome must stay awake and running.
How to start a task in Gemini, step by step
- Open Gemini. On the web or the Mac app, look for Task below the text box. On Android it sits above the text box. If you do not see it, check the requirements above first.
- Click or tap Task. The first time, follow the prompts to turn task mode on.
- Describe the goal, not the clicks. Say what finished looks like, the limits (budget, dates, which folders or labels to touch) and how you want the result delivered, for example as a Google Doc or a spreadsheet. If it should run later or repeatedly, say so in the same message: a specific time, or an event such as "when the price drops below".
- Add context. Type @ to pick one of your skills, and use Upload & tools to add files (on Android the options are Upload files, Add from Drive and Add Notebook). You can change the model from the model menu at the top.
- Click Submit. The task gets its own thread in the sidebar (on Android, under Menu). You can add instructions while it runs without losing progress.
- Watch the work panel. Click the progress chip at the top of the thread to see completed, current and planned steps. The Files section lists what Gemini read or changed. Check back regularly, because Gemini may pause and wait for your input.
- Approve or take over. The first time Gemini browses on a computer, Chrome asks you to connect Chrome auto browse; after that Gemini asks before each browsing task. It always asks before sensitive actions: sending messages, changing or deleting data, purchases and other financial transactions, submitting forms and signing in with Google. For a remote browser, click View live, then Take over task (on Android, Take control), and hand it back with Go back to Gemini (on Android, End control).
- Stop it if needed. Stop in the Gemini in Chrome side panel halts browsing only; Gemini may then try another route or give up. Stop in the Gemini text box cancels the whole task.
- Manage schedules. Open the task thread, click the scheduled task chip, then More options and Pause or Resume. On the web there is also a Scheduled page.
Two limits matter in practice. You can run up to 15 tasks at once, and a scheduled task will not start if 15 are already running. Usage limits depend on your plan and on how complex the task is, and Google does not publish exact numbers.
A worked example: a flight price watch that ends in a decision
Say you need to be in Lisbon for a conference in March and your budget is fixed. With the toggle set to Task, you might write: "Watch return flights from Chicago to Lisbon for 9 to 13 March. My budget is 900 dollars including one cabin bag. Check daily. When a fare is under budget, tell me in this thread with the airline, times, total price and the booking link. Do not book anything."
Here is what that wording does. The dates, airport and budget give Gemini a test it can check without you. "Check daily" makes it a scheduled task, so it shows a scheduled task chip you can pause later. The delivery line tells it what a useful result looks like. "Do not book anything" is belt and braces: Google says purchases always need your approval anyway, but saying it removes any doubt about intent.
While it runs, open the thread and click the progress chip to see which sites it checked. If it hits a page that needs you to sign in, the task pauses until you take over. When the alert arrives, book it yourself on the airline's site. Once you have booked, open the thread and pause the schedule so it stops using your allowance.
The same pattern works for Google's inbox example. Ask Gemini to list every newsletter you received in the last 30 days, sorted by how often you open them, and to archive the ones you have not opened, then stop and show you the unsubscribe list before acting on it. You get the tedious sorting done and keep the final call.
Common mistakes and how to fix them
- The Task option is missing. The usual causes are a work or school account, a Google AI Plus or free plan, being in the UK, EEA, Switzerland or Nigeria, or using the Windows app. Switch to a personal account on Pro or Ultra in a supported country, on the web, Mac or mobile app.
- Typing passwords or card numbers into the thread. Google tells you not to. Enter sign-in details and payment information directly on the website when Gemini hands you control.
- Expecting a local Chrome task to run overnight. Local browsing needs your computer and Chrome awake. If the device is off, Gemini might switch to the remote browser, which cannot see your logged-in sites, so the task stalls at the first sign-in.
- Vague goals. "Sort my inbox" gives Gemini nothing to test against. Name the labels, the date range, what to archive and what to leave alone.
- Trusting web content blindly. Google warns about prompt injection: instructions hidden in a web page or email can try to steer the agent. Keep sensitive tasks narrow and read every approval request before you accept it.
- Confusing a task schedule with a chat schedule. If you want a simple recurring summary, a chat schedule (up to 10 active, managed from Scheduled in the sidebar) is enough. Use a task when the job needs several steps or actions.
- Turning task mode off to tidy up. In Settings > Task settings, Turn off under "Use task mode to hand off work" deletes your remote browser and code execution data and pauses your schedules, which do not run while task mode is off. Your tasks, schedules and files are kept. To clear only remote browsing data, use Settings > Browser > Delete remote browser data instead.
Gemini Tasks, ChatGPT Work or Claude Cowork: which to use
All three let you hand over a multi-step job rather than ask a question, but they suit different people.
Pick Gemini Tasks if your work lives in a personal Google account: Gmail, Calendar, Drive, Docs, Sheets, Keep and Google Tasks are all in reach, plus YouTube, Photos and Google's search services. Event-triggered schedules, such as price alerts, are a real strength. It is not for you if you need it on a work account, live in the UK or EEA, or are on the free plan.
Pick ChatGPT Work, the part of OpenAI's ChatGPT built for multi-step jobs, if you want it to build a website or a presentation or do things on the web from your files and plugins. In an email reported by TechCrunch and 9to5Mac on 2 November 2026, OpenAI said it will start testing a combined Chat and Work experience with a small group of users in the coming weeks. Check the ChatGPT pricing page for what your plan includes.
Pick Claude Cowork if you work across Google Drive, Gmail, Microsoft 365 and Slack, or need an agent on a work plan. In Anthropic's Claude apps, Cowork is now part of every conversation, rolling out first to Pro, Max and Ultra, with Team plans in beta. Cloud and scheduled tasks keep running when your computer is off; tasks that use local files need Claude Desktop open. See ChatGPT vs Gemini and our guide to using AI agents for the wider picture.
If your team is waiting for an agent on Google Workspace, the Gemini agent that Google Cloud announced at Gemini at Work is the product to watch, not Tasks. Meanwhile, writing your repeat jobs as skills pays off either way; our guide to moving Gemini Gems to skills shows how, and Learn Gemini covers the basics.
Prompts to copy
Inbox declutter with a checkpoint
Find every newsletter and mailing list email in my Gmail from the last [30] days. Make a table of sender, how many emails, and whether I opened any. Archive the ones I never opened. Then stop and show me the list you would unsubscribe from. Do not unsubscribe until I approve.
You get: A cleaner inbox and a list to approve before Gemini acts
Price watch that never buys
Check [flight or hotel] prices for [route or city] on [dates] every [day]. My budget is [amount] in total. When something is under budget, post the option, total price and link in this thread. Do not book or enter any payment details.
You get: A scheduled alert you can act on yourself
Research packed into a file
Research [topic] using sources from the last 12 months. Put the findings in a Google Doc with a one-paragraph summary, the five most important facts with a source link for each, and open questions at the end. Tell me which sources you could not open.
You get: A downloadable research file with checkable sources
Post-holiday catch-up
I was away from [start date] to [end date]. Read my Gmail and Google Calendar for that period and make a to-do list in Google Tasks: replies I owe, deadlines in the next two weeks, and meetings I need to reschedule. Do not send any emails.
You get: A prioritised to-do list without anything sent on your behalf
Weekly digest that follows a story
Every [Monday at 8 a.m.], give me a digest of news on [topic or company]. For each story, say what changed since last week's digest and link the original source. Keep it under [300] words.
You get: A recurring digest that tracks how stories develop