⚡ Quick verdict

We may earn a commission if you buy through our links, at no extra cost to you.

For six months, Google's most capable AI agent was a paid extra. Spark — later renamed Tasks — launched in May 2026 for Google AI Pro and Ultra subscribers, which meant $19.99 a month at minimum to have an AI that would do things for you instead of just answering questions. On 17 November 2026, Google removed the paywall. Tasks in Gemini is now free in six countries, running on the new Gemini 4 Hydrogen model, and the US version arrives with three abilities no free Google agent has had before: it can phone a business, it can pay for something, and it can sign in to a website on your behalf.

That is a real shift, not a rebrand. A chatbot answers when you ask. An agent works while you do something else — researching a purchase, watching a topic, calling two dealerships at once to compare service prices, then putting the appointment on your calendar. Google's announcement, posted 17 November 2026 by Josh Woodward, vice president of Labs, the Gemini app and AI Studio, frames it plainly: put Gemini to work for you. The question this guide answers is what "free" actually covers, what is still locked behind a subscription, where the new powers stop at the US border, and whether you should trust an agent with your calendar, your money and your passwords.

One thing to say plainly up front, because the word "free" does a lot of work in the headlines: a Google account is required. Tasks in Gemini is free in the sense that it costs no money in the six launch countries. It is not no-signup, it is not anonymous, and it does not work with a work or school account. If that distinction matters to you, it matters before you read any further. For the wider chatbot landscape, our best AI chatbot roundup compares Gemini with ChatGPT, Claude and the rest — and our best AI coding assistants guide covers the developer side of the same agent wave.

Try Gemini Tasks →

What is Gemini Tasks?

Gemini Tasks is Google's always-on AI agent inside the Gemini app. The difference between an agent and a chatbot is the difference between an assistant you talk to and an assistant you delegate to. A chatbot waits for your next message. An agent takes a job, goes away, and comes back with a result — or a progress report, or an admission that it got stuck.

Here is how it works in practice. You flip the Gemini home-screen toggle from Chat to Tasks, and describe the job in plain language: "research the best robot vacuum under $300 and send me a comparison by Friday," or "watch for news about the new Pixel and tell me when a review lands," or "book my car in for a service next week." Gemini works on the job in the background, in what Google describes as dedicated, hardware-isolated virtual machines, and notifies you when it has a result. You do not sit and watch it browse; you get on with your day.

The lineage matters, because it explains why this is news. Google announced Spark on 19 May 2026 as a 24/7 agent for paying subscribers: give it a job and a trigger — every Monday at 7, or whenever a certain email arrives — and it does the work in the background, including in a browser. For six months it stayed behind the Pro and Ultra paywall. On 17 November 2026 Google renamed it Tasks and made it free in six countries, on the new Gemini 4 Hydrogen model — a model Google introduced on 12 November 2026, built for rapid execution across agentic workflows. The rename is why you may not find "Spark" anywhere in the app any more: it is now simply Tasks in Gemini.

What can it actually do? The honest answer is: a lot of useful things, inside Google's world, and a growing list outside it. Research a topic and return a report. Monitor a subject and send updates on a schedule. Plan a trip across flights, hotels and activities. Compare products across stores. Draft and organise from your email. And, in the US, the three new powers that make this launch different from every previous free Gemini update: phone calls to businesses, checkout with Google Pay, and sign-in through Google Password Manager. Each of those is covered in detail below, because each has limits that the headlines skip.

What is free now — and what still costs money

The free tier is generous, but it has a shape. Understanding the shape now saves a surprise later.

Free in six countries. Tasks in Gemini is free of charge in the US, Japan, Korea, Canada, Australia and India. Google describes the rollout as reaching hundreds of millions of people. Outside those six countries, Tasks still needs a Google AI Pro or Ultra plan, and it is not offered at all in the European Economic Area, Nigeria, Switzerland or the UK. If you are reading this from London or Berlin, the free tier does not apply to you yet — check Google's availability pages for your country rather than assuming the launch covers you.

Schedules: once a day on free. The free version — and, according to one detailed tracker, the $4.99 AI Plus tier — can schedule a task up to once a day, in a general window: overnight, daytime or evening. You can have Gemini check something daily, but you cannot have it check hourly, and you cannot set a precise time. Hourly schedules and Gmail monitors and news-topic monitors still need a Pro or Ultra plan. That is the real paywall now: not access to the agent, but how often it runs and what triggers it. For most personal use — a daily news digest, a daily price check, a weekly research roundup — once a day is enough. For a business that wants an agent watching an inbox all day, it is not.

The core agent is free. Research, planning, comparisons, background execution and notifications are all included in the free tier in the six countries. There is no separate "agent credits" system to buy into, and no per-task charge has been announced. Google AI Pro ($19.99 a month) and Ultra add the scheduling frequency, the Gmail and news triggers, and access in countries where the free tier is not offered — plus the wider Pro and Ultra benefits across Gemini, such as higher limits elsewhere in the app. If you were already paying for Pro to get Spark, you are now paying for frequency and triggers rather than for the agent itself.

The US-only extras are free too — where they exist. Phone calls to businesses, Google Pay checkout and password-manager sign-in are included in the free tier in the US, with no extra subscription. They are limited by geography, not by plan — a Pro subscriber in Japan does not get US calling, and a free user in the US does. That is worth knowing if you are weighing a Pro subscription: it does not buy you the calling feature abroad.

The bottom line on price: for a personal user in one of the six countries, Tasks costs nothing beyond the Google account you already have. The costs that remain are the ones Google has always charged for — the Pro subscription, for power users who need hourly runs — and the costs of the things the agent buys on your behalf, which are governed by the spending cap you set.

The US-only extras: calls, payments and sign-in

The three new capabilities are the reason this launch matters beyond the price cut, and they are also the reason to read the fine print. All three are rolling out in the US only, for now.

Phone calls to local businesses. Gemini can find nearby businesses on Google Maps and call them, in parallel, to get prices, check inventory or book appointments. Google's own example is a car service: Gemini used Maps to find nearby dealerships, called two of them simultaneously, checked the user's Calendar for availability, and booked the appointment. That is the Duplex lineage — Google's phone-calling AI, first shown years ago — now wired into an agent that also knows your schedule. The practical uses are obvious: booking a table, checking whether a shop has an item in stock, comparing quotes from two garages, making a hair appointment without sitting on hold.

There are three limits to know. First, geography: calling is not currently available in Indiana, Louisiana, Minnesota, Montana and Nebraska, and the feature only calls US phone numbers — it cannot call a business abroad. Second, disclosure: Google says Gemini discloses right away that it is an AI, at the start of the call, and businesses can opt out of such calls. Google documents two different opt-out routes and has not said clearly which one governs Tasks, so a business that wants out may need to check both — the Gemini calling opt-out page and the general automated-calls opt-out. Third, the calls are for a customer's errand, not for cold outreach: Gemini calls on your behalf about your booking or your question, and Google's help pages frame the feature that way.

Checkout with Google Pay. Gemini can compare products and check out with Google Pay, using the Agent Payments Protocol (AP2), with Mastercard and Visa supported on eligible cards — American Express support is reported as limited or absent, so check the official pages if you pay with Amex. You set a spending cap, and the agent works inside it. The idea is that the agent does not just find the product; it completes the purchase. That is genuinely new for a free Google product, and it is also the feature that most needs a low cap while you learn what the agent does with it. Set the cap low, watch the first few purchases closely, and only raise it once you trust the judgement.

Password-manager sign-in. Gemini can sign in to sites using accounts stored in Google Password Manager, without sharing your password with the site in the usual way. The point is to let the agent act inside a logged-in account — checking an order, managing a booking, updating a profile — without you handing over credentials or staying logged in yourself. It is the capability that makes the other two work end to end: the agent can call the business, then log in to the booking site, then pay. It is also the capability that requires the most trust, and it is covered properly in the privacy section below.

Two new ways in: text and email. In the US, you can also reach Tasks by texting Gemini or by forwarding an email to gemini@google.com. Texting requires a US phone number on an eligible mobile carrier. Forwarding an email is the simplest on-ramp of all: send the thing you want handled to the agent, and it takes it from there. Both are US-only at launch, and both are worth trying before you decide whether the toggle-based workflow suits you.

Requirements: who can actually use it

Free does not mean open to everyone, and the requirements are specific enough to check before you get excited.

A personal Google account. Google's help page says the free version requires a personal Google account. Work and school accounts are excluded for now — if your only Google account is a Workspace account from your employer or university, you cannot use Tasks yet. That exclusion is easy to miss and it is the single most common reason someone will open Gemini, look for the toggle, and not find it.

18 or older, with Gemini history on. The product page limits free access to users aged 18 and over, and Google's help page says Gemini history must be turned on. History is what lets the agent remember your jobs and your preferences across sessions — without it, the agent cannot do the background work that defines it. If you have history off for privacy reasons, you face a real trade: the agent's usefulness against the record it keeps.

One of the six countries. The US, Japan, Korea, Canada, Australia and India. Pro and Ultra subscribers in selected other countries also have access, but the free tier does not travel. And three regions are explicitly not offered: the European Economic Area, Nigeria, Switzerland and the UK. If you are in one of those, no plan currently buys you Tasks — check Google's availability pages for updates rather than assuming a launch is coming soon.

US-only extras need the US. The calls, the Google Pay checkout and the password-manager sign-in are US-only, and calling excludes five states: Indiana, Louisiana, Minnesota, Montana and Nebraska. Texting Gemini needs a US phone number on an eligible carrier. If you are in Canada, Australia, India, Japan or Korea, you get the free agent — research, monitoring, planning — but not the phone calls or the checkout. That gap is likely to close over time, but Google has not given a date.

How to try Gemini Tasks

Getting started takes about two minutes, assuming you meet the requirements above.

Step 1: open Gemini. Go to gemini.google.com in a browser, or open the Gemini app on your phone. Sign in with a personal Google account — not a work or school account — and make sure you are 18 or older and have Gemini history turned on. If the Tasks toggle is not there, check the country and account requirements first; that is the cause almost every time.

Step 2: flip the toggle. On the Gemini home screen, switch the control from Chat to Tasks. That one toggle is the whole interface change: Chat is the chatbot you already know, Tasks is the agent. If you have used Spark before, this is the same feature under its new name — your existing scheduled tasks should carry over, though it is worth checking they survived the rename.

Step 3: give it a low-stakes job first. Do not start with a payment or a password. Start with something you can check in thirty seconds: "Research the best wireless earbuds under $100 and send me a comparison by tomorrow evening," or "Watch for news about the next iPhone and tell me when a major review is published." You want to learn three things from the first job: how long it takes, how good the result is, and whether the notification actually arrives.

Step 4: try a scheduled task. The free tier allows schedules up to once a day, in a general window — overnight, daytime or evening. Set a daily job you genuinely want: a morning news digest on a topic you follow, a daily check on a price you are watching, a weekly summary of a subject. This is where the agent earns its keep: the value is not in one clever answer, it is in the job that happens every day without you asking.

Step 5: try the US extras — carefully. If you are in the US, try a phone call for something low-stakes: call a restaurant to check opening hours, or two shops to compare a price. Set a low spending cap before you try a purchase, and try the password-manager sign-in on a site where a mistake costs nothing. Check every result. Google itself cautions that tasks can still fail to complete or miss their objective, and the first week is when you learn where the edges are.

Step 6 (US only): try text or email. Text Gemini, or forward an email to gemini@google.com, and see whether the agent picks the job up. Forwarding is the most natural way to delegate the kind of errand that arrives in your inbox — a booking confirmation to follow up, a bill to check, a reservation to change — and it is worth testing with something you can verify.

What it does well — and where it fails

Where it shines. The strongest use cases are the ones that combine research with waiting. Monitoring a topic and reporting back. Comparing options across many sources and returning a shortlist. Planning a multi-step errand — find the business, check the price, check the calendar, book the slot — where each step is simple and the value is in doing all of them without you. The background execution is the point: you ask once, and the work happens while you do something else. For anyone who has ever kept a browser tab open for three days meaning to compare two products, that is a genuine improvement.

Where it struggles. Google itself concedes the feature still fails in places, and the honest reading is that agentic AI in November 2026 is good at well-defined jobs and shaky at ambiguous ones. A job with a clear end state — "book the cheapest appointment next Tuesday" — is easy to verify. A job with a judgement call — "find me a good restaurant for a date" — depends on what "good" means to you, and the agent will guess. Jobs that cross many unrelated websites are harder than jobs inside Google services, because the agent has to navigate logins, pop-ups and layouts it was not built for. And any job where a mistake costs money — a wrong booking, a wrong purchase — needs you to check the result, which eats into the time the agent was supposed to save.

The notification matters more than it sounds. An agent that works in the background is only useful if it tells you when it is done, and tells you clearly what it did. Pay attention to the first few notifications: do they arrive promptly, do they say what was actually booked or bought, do they flag the cases where the task failed? If the reporting is vague, treat the agent as a research assistant rather than a delegate — useful for finding things, not for finishing things.

The once-a-day limit shapes the use cases. A daily schedule is enough for news, prices and roundups. It is not enough for anything that needs watching through the day — a flash sale, a breaking story, an inbox that needs triage. That is the deliberate shape of the free tier: useful enough to keep, limited enough that heavy users still have a reason to pay. If you find yourself wanting hourly checks or Gmail triggers, that is the moment Pro earns its $19.99 — not before.

Gemini Tasks vs ChatGPT agent mode vs Claude

Gemini Tasks is not the only agent in town, and the choice between them depends on where your errands live. Here is the honest comparison as of November 2026.

ChatGPT agent mode. OpenAI's agent is the strongest pick for general computer work in a browser: filling in forms, working across many different websites, chaining tools together in an open-ended session. It is available in more countries than Gemini Tasks, and its limits are better documented from months of public use. What it cannot do as deeply is the Google-native work: it does not have your Calendar, it does not call businesses from Maps, it does not pay with Google Pay, and it does not sign in with Google Password Manager. If your errands are mostly calls, bookings and Google services, Gemini Tasks is the better tool — and it is free. If they are mostly open-web browsing across sites Google does not own, ChatGPT agent mode is still ahead. Our best AI chatbot roundup compares the underlying models in detail.

Claude. Anthropic's Claude is the strongest pick for careful, long-form work — research, writing, analysis, and coding help where accuracy matters more than speed. Claude's agent features are aimed more at developers and knowledge workers than at everyday errands: it will not call your local garage or book a restaurant. If your "tasks" are really "think hard about this document," Claude is the better choice. If your tasks are "book the thing and pay for it," Gemini Tasks is. The two are complements more than competitors, and many people will end up using both.

The coding angle. For developers, the agent wave is also reshaping coding workflows — agents that write, test and fix code in the background. Gemini Tasks is not a coding agent, and it does not try to be. If that is the job, our best AI coding assistants guide covers the tools built for it. The point is to match the agent to the job: a general-purpose errand agent for errands, a coding agent for code.

The sales angle. One group worth a specific mention is small sales teams. An agent that monitors an inbox, follows up on a lead, checks a calendar and books a call is doing the first half of an SDR's job — and the tools built specifically for that work, covered in our best AI SDR tools roundup, are still better at it than a general agent. Gemini Tasks is a personal errand agent first; the specialised tools win when the errand is a pipeline.

Who Gemini Tasks is for — and who should skip it

Busy people in the six countries. If you live in the US, Japan, Korea, Canada, Australia or India, have a personal Google account, and your week is full of small errands — bookings, comparisons, follow-ups, research — Tasks is worth an afternoon of testing. The price is zero and the setup is a toggle. The value shows up in the third or fourth delegated job, not the first.

US users who hate phone calls. The calling feature is the single best reason to try it. If you have ever put off booking an appointment because it meant a phone call, an agent that calls two businesses at once and puts the result on your calendar is a real quality-of-life improvement. Start with a low-stakes call and check the result.

Google households. The more of your life is already in Google services — Calendar, Maps, Pay, Password Manager, Gmail — the more the agent can do without you. If your life is in Apple or Microsoft services instead, the agent has less to work with, and the value drops accordingly.

Who should skip it. Skip it for now if you are outside the six countries, if you only have a work or school account, or if you are under 18 — the feature simply is not available to you yet. Skip the payment and sign-in features until you have tested the agent on free, low-stakes jobs and trust its reporting. And skip it if you are uncomfortable with the data bargain: the agent needs your history on, your calendar, your location and your payment method to do its job, and that is a lot of access to hand to any system.

Gemini Tasks

Strengths

Weaknesses

Try Gemini Tasks →

Privacy and trust: read this before you switch it on

An agent that can call, pay and sign in needs more access than a chatbot, and the trust question is the most important one in this guide. Here is what we know, stated plainly.

What Google says about how it runs. Google says tasks run in dedicated, hardware-isolated virtual machines — separate, sealed-off computing environments for each job. That is a meaningful safeguard: the agent's browsing and execution happen away from your main session. Google also says Gemini discloses right away that it is an AI on phone calls, and that businesses can opt out of the calls. Purchases use the Agent Payments Protocol with digital-token support through Mastercard and Visa on eligible cards, and you set a spending cap.

What you are handing over. To do its job, the agent needs your Gemini history on, access to your Calendar, your location (to find nearby businesses), your payment method, and your stored passwords. That is the bargain, and it is worth making consciously. History on means Google keeps a record of your tasks and conversations — if you have kept history off for privacy reasons until now, switching it on is a real change, not a checkbox.

The failure mode to plan for. Google cautions that tasks can still fail to complete or miss their objective. The realistic failure is not a dramatic one — it is a quiet one: an appointment booked at the wrong time, a purchase of the wrong variant, a call that reached the wrong branch. The defence is boring and effective: set a low spending cap, start with low-stakes jobs, and check every result before you trust the agent with anything that costs real money or cannot be undone.

Businesses and the opt-out. If you run a business, the other side of the calling feature matters to you: Gemini may soon be calling your shop on a customer's behalf. Google says businesses can opt out whenever they want, but its own documentation is unclear about which opt-out governs Tasks — Google documents two routes, the Gemini calling opt-out page and the general automated-calls opt-out, without saying which one applies. If you want out, check both, and verify the number is actually excluded rather than assuming the form worked.

The honest summary. Google has built more safeguards into Tasks than most agent launches get — isolated VMs, AI disclosure on calls, a spending cap, an opt-out for businesses. Those are real. But no safeguard removes the need for you to check the work. Treat Gemini Tasks like a capable new employee: give it clear jobs, check the first month of results carefully, and only widen its authority once it has earned it.

Our verdict

Gemini Tasks going free is the most consequential AI launch of the week, and the reason is not the price — it is the powers. A free agent that can call a business, book the appointment and pay for it is a different product from a free chatbot, and Google is the only company that could build it, because only Google owns the Maps, the Calendar, the Pay wallet and the password manager in one account. For a personal user in the US, the combination is genuinely useful from day one.

But the honest verdict has limits, and they are geographic. Outside the US, you get the free agent without the three powers that make it special — research, monitoring and planning, but no calls, no checkout, no sign-in. Outside the six countries, you get nothing free at all. And everywhere, the agent is young: Google itself says tasks can fail, the free schedule is once a day, and the trust question — your calendar, your money, your passwords — is one only you can answer.

Our recommendation is simple. If you are in one of the six countries with a personal Google account, flip the toggle and give it three low-stakes jobs this week: a research task, a daily monitor, and — if you are in the US — one phone call. Check the results, watch the notifications, and keep the spending cap low. If the three jobs come back right, you have just hired a free assistant. If they do not, you have spent nothing and learned exactly where the edges are. Either way, the agent wave is no longer a paid preview — it is free, it is in your pocket, and it is worth an afternoon of your time.

Frequently asked questions

Is Gemini Tasks really free?

Yes — since 17 November 2026, Tasks in Gemini is free in the US, Japan, Korea, Canada, Australia and India, running on Google's new Gemini 4 Hydrogen model. It used to require a paid Google AI Pro subscription — $19.99 a month — under its old name, Spark. The free version covers scheduled tasks up to once a day, in a general window such as overnight, daytime or evening. Hourly schedules and Gmail and news triggers still need a Pro or Ultra plan. A personal Google account is required: the feature is free, not no-signup.

What can Gemini Tasks actually do?

It is an agent, not a chatbot: you give it a job and it works on it in the background, then notifies you when it is done. It can research a topic and return a report, monitor a subject and send you updates, plan a trip, compare products, and handle multi-step errands across Google services. In the US it can also find local businesses on Google Maps, call them to check prices or stock or to book an appointment, check out with Google Pay, and sign in to sites using Google Password Manager. You switch it on by flipping the Gemini home-screen toggle from Chat to Tasks.

Can Gemini Tasks call businesses for me?

Yes, in the US only — and not everywhere in the US. Gemini finds nearby businesses on Google Maps and can call more than one at the same time, for example two dealerships, to compare prices or book a service appointment, then adds the booking to your Google Calendar. Google says Gemini discloses right away that it is an AI, and businesses can opt out of the calls. Calling is not currently available in Indiana, Louisiana, Minnesota, Montana and Nebraska, and the feature only calls US phone numbers.

Do I need a Google account to use Gemini Tasks?

Yes. Tasks in Gemini requires a personal Google account, and Google says you must be 18 or older with Gemini history turned on. Work and school Google accounts are excluded for now, so if you only have a Workspace account you cannot use it yet. Outside the six free countries — the US, Japan, Korea, Canada, Australia and India — Tasks still needs a Pro or Ultra plan, and it is not offered in the European Economic Area, Nigeria, Switzerland or the UK.

Is Gemini Tasks better than ChatGPT agent mode?

It depends on what you want the agent to do. Gemini Tasks is the stronger pick if you live inside Google services: it can check your Calendar, call businesses from Maps, pay with Google Pay and sign in with Password Manager — things ChatGPT agent mode cannot do as deeply. ChatGPT agent mode is the stronger pick for general computer work in a browser, such as filling forms or working across many different websites, and it is available in more countries. If your errands are mostly calls, bookings and Google services, try Gemini Tasks first — it is free. If they are mostly open-web browsing tasks, ChatGPT agent mode is still ahead.

Can I trust Gemini Tasks with payments and passwords?

That is the right question to ask before you switch it on. Purchases go through Google Pay using the Agent Payments Protocol, with Mastercard and Visa supported on eligible cards, and Google says you set a spending cap. Password-manager sign-in lets Gemini log in to sites without sharing your password with the site directly. Google also runs tasks in dedicated, hardware-isolated virtual machines and says it discloses its AI identity on calls. Even so, Google itself cautions that tasks can still fail to complete or miss their objective — so set a low spending cap, start with low-stakes errands, and check the result before you trust it with anything that costs real money.

AT
About Aitopkit

We research AI tools in depth before recommending them — official pricing, public docs, and real user reports. If you buy through our links we may earn a commission at no extra cost to you. Read our disclosure.