Google today made the task mode in its Gemini app free for what it described as "hundreds of millions of people" in the US, Japan, South Korea, Canada, Australia and India, five weeks after introducing the feature in a post that named only paying Google AI Pro and Ultra subscribers. The free rollout runs on Gemini 4 Hydrogen. A second set of features, starting in the US, lets Gemini place phone calls to businesses, pay with tokenized cards through the Agent Payments Protocol, take instructions by text message and act on forwarded emails.
In Short
Google's Gemini app can now carry out errands such as booking a car service by phone, digging through old utility bills or buying a charger, and Google today opened that task mode to free users in six countries. The change puts an AI agent that phones businesses, signs into websites with stored passwords and pays with tokenized cards in front of an app with more than 1 billion monthly users, so local businesses, retailers and website operators may increasingly deal with Gemini rather than a person. The calling, texting, email and payment features start in the US, calling is unavailable in five states, and Google says the agent still makes mistakes and sometimes hands the final click back to the user.
What Google announced
The announcement came as a first-person post on Google's Keyword blog, published at 17:00 UTC and filed under Gemini App, AI and Google Labs. Its author is Josh Woodward, Vice President of Labs, the Gemini app and AI Studio. The headline reads "Put Gemini to work for you"; the browser title is "How to use Tasks in Gemini for personal assistance." Google lists it as a six-minute read. The subheading sets out the ambition plainly: "Gemini can now make the calls, pay the bills, and finish your tasks."
Two sentences carry the commercial news. "Starting now, this update is rolling out free of charge for hundreds of millions of people across the U.S., Japan, Korea, Canada, Australia, and India, using our new Gemini 4 Hydrogen model," the post states. Then: "We're also rolling out a set of advanced features, starting in the U.S."
The post never itemizes that second set. Its footnotes and closing instructions fill some of the gap. Calling is restricted by US state, texting requires a US phone number, and email forwarding is introduced with the words "Starting in the U.S." For payments, no geography is given at all.
The rest of the post is a sequence of household errands from Woodward's own life, used to show Gemini working as a personal AI agent that holds the user's context and acts on it. "The best assistant knows you and helps you get things done," it opens. "Our goal is for Gemini to save you time and money."
Google published a companion post at the same minute. Written by P.J. Linarducci, VP of Consumer Payments, it announced that Google Wallet is now available on iOS in the US, where it gathers receipts, package tracking, boarding passes and tickets pulled from Gmail and other Google apps. That post notes that tap-to-pay is not supported on iOS. Gemini's purchase receipts are designed to land in the same wallet, which is why the two announcements arrived together.
What is available, and where
| Feature | Availability stated by Google | Condition or limit stated |
|---|---|---|
| Task mode (home-screen toggle from Chat to Task) | Free in the US, Japan, Korea, Canada, Australia and India | Runs on Gemini 4 Hydrogen |
| "Advanced features" | Starting in the US | Not itemized |
| Phone calls to businesses | US | "not currently available in Indiana, Louisiana, Minnesota, Montana, and Nebraska" |
| Text with Gemini (RCS) | US | "You must have a US phone number with an eligible mobile carrier" |
| Forwarding emails to gemini@google.com | Starting in the US | None stated |
| Purchases through the Agent Payments Protocol (AP2) | Not stated | Mastercard and Visa; American Express "coming soon"; digital tokens on supported cards |
| Receipts and tracking in Google Wallet | Wallet expands to iOS today (US, per the companion post) | Tap-to-pay not supported on iOS |
| Connected apps | Not stated | "more than 60 popular apps" |
| Universal Commerce Protocol in Gemini | Not yet available | Described as "Up next," no date |
From $19.99 to free in five weeks
The speed of the shift is the story. On October 13, PPC Land reported that Google had added a toggle from chat to tasks in the Gemini app, released alongside Gemini 4 Argon, in a post whose only availability sentence named Google AI Pro and Ultra subscribers. That post said nothing about free users or the AI Plus tier. Google's plan documentation, as reported by PPC Land on October 4, put AI Plus at $4.99 a month, AI Pro at $19.99 and AI Ultra at $99.99 or $199.99, and limited accounts without a subscription to Flash-Lite from October 9.
Today, 35 days later, the same toggle is described as free. The wording has shifted slightly, from "chat to tasks" in October to "Chat to Task" now, but the mechanism is the same switch on the app's home screen.
The model has changed, too. Today's post does not mention Argon. It names Gemini 4 Hydrogen, which Google introduced on November 12 as a faster and cheaper model for agentic work. PPC Land reported that day that Hydrogen had reached Google Search for all users, with developer pricing of $1.50 per million input tokens and $7.50 per million output tokens, and that Google's model post said free Gemini app users would reach Hydrogen only through automatic model selection. Today's post names Hydrogen as the model for the free task rollout. It does not say whether paying subscribers' tasks still run on Argon, or whether free tasks face usage limits.
Google has been moving agent features down its price ladder for several weeks. When PPC Land covered a Google holiday guide on November 16, the help page for AI Mode's monitoring tasks named no paid tier, although Google had said at I/O in May that those agents would launch first for AI Pro and Ultra. On November 10, Google extended Android task automation, which works through food and ride apps, to the UK, India and Japan, with no subscription required according to its help page.
The country list deserves a second look. It includes no European country. The UK, which received Android task automation a week ago, is absent from today's list. So is every member of the European Economic Area. That pattern is not new: PPC Land reported on July 30 that Gemini Spark, Google's background agent, was unavailable in the EEA, the UK, Switzerland and Nigeria even as it expanded to more than 160 other countries. Today's post gives no reason for the selection and no timetable for other markets.
Five errands, as Google tells them
The post demonstrates the product through five tasks. Each relies on a different combination of Google services, and together they show how much of a household's administration Google now proposes to route through one agent.
| Errand | What Gemini did, according to the post | Google services and partners named | Figure given |
|---|---|---|---|
| Car service | Called the two nearest dealerships at once, booked an early Friday appointment, added it to the calendar | Personal Intelligence, Google Maps, Google Calendar | "less than five minutes" |
| Electric bill | Signed into the utility portal, analyzed past statements in email, found the heat pump running around 4 p.m. during peak pricing | Google Password Manager, Gmail | $389 bill; "a few minutes" of work |
| USB-C charger | Compared chargers, asked about wattage and color, checked out, found a discount code | Shopping Graph, Google Pay, AP2 | $10 saved |
| Groceries | Added items to an Instacart order from a text message and a photo, then handed off to Instacart | RCS messaging, Instacart | None |
| School schedule | Put every date from forwarded school emails on the calendar, invited Woodward's wife, flagged a clash with basketball practice | Gmail, gemini@google.com, Google Calendar | "less than a minute" |
The phone call
The first example is the most consequential for local businesses. "Because Gemini already knows we drive a 2019 Honda Odyssey and where we live through Personal Intelligence, it found the nearest shops on Google Maps and called both at once," Woodward writes. An image caption adds that Gemini "booked an appointment at the shop that accepts appointments vs. walk-ins only."
On disclosure, the post is specific. "On the dealership's end, Gemini disclosed right away that it's an AI. Businesses can opt out whenever they want, and Gemini won't give out my payment or private account info unless I say so."
The footnote attached to the call reads: "Calling through Gemini is not currently available in Indiana, Louisiana, Minnesota, Montana, and Nebraska." The same five states were excluded when Google introduced AI-powered business calling in US Search on July 16, 2025, a feature that checked prices and availability on a user's behalf. The repetition suggests the exclusions follow from the same underlying calling system, though Google does not say why those states are left out.
Gemini is now the third Google consumer surface that phones businesses. In November 2025, Google launched "Let Google Call" to check stock and prices at nearby stores, alongside agentic checkout. Yesterday, PPC Land reported that Ask Maps had started phoning salons, pet groomers and auto and home repair businesses to find open appointments, starting in the US.
The opt-out link in today's post leads to a Google Business Profile help page titled "About automated calls and texts from Google to your business." That page does not mention Gemini or AI by name. It says businesses can tell Google over the phone or change their Business Profile settings, under Advanced settings and a section called "Google automated calls and text messages," which covers bookings and inquiries from customers, profile updates and posts made on the business's behalf. It also warns that a business that opts out may still receive occasional manual calls from Google to confirm its information.
The bill
The second example introduces a capability that has received less attention: signing into a third-party website with the user's stored credentials. "I gave Gemini permission to access our utility portal with the new Google Password Manager integration and asked for it to analyze our past billing statements in my email," the post says. The analysis found that the family's heat pump was "kicking on around 4 p.m., right in the middle of peak pricing."
The post does not explain how a password moves from Password Manager into the agent's environment, whether the model ever sees it in plain text, or what happens if a portal asks for a second authentication factor. Stored credentials are not new territory for Google's agents: Chrome auto browse, available on desktop in the US, can use saved logins and passwords but hands payment steps back to the user.
The charger
For shopping, the post leans on Google's Shopping Graph, the product data set behind its shopping results. Gemini "quickly compared chargers thanks to Google's Shopping Graph, asked me follow-up questions about the wattage and color, and then helped me check out using the new Google Pay integration," according to Woodward. "The best part: it also found a discount code so I saved an extra $10."
That last detail matters for anyone who distributes promotional codes. A shopping agent that finds and applies codes on its own changes where codes are found and who gets credit for them. When Universal Cart reached US Search and the Gemini app on October 15, it already displayed promo codes and Google Wallet card perks. Today's post does not say where Gemini finds codes, or whether merchants can control which ones it applies.
Groceries by text
The fourth example uses a new channel. Woodward texted Gemini "with the new RCS integration," asking it to add lasagna, strawberries, cucumbers and vanilla ice cream to an Instacart order, and later sent a photo of an empty cereal box with the words "Add this too." Gemini "added it, and then handed me off to Instacart to complete the order."
The handoff is significant. The user, not Gemini, completed the order inside Instacart's own flow, so Instacart kept the checkout. Google says Gemini "connects with more than 60 popular apps," and invites developers to apply to its Connected Apps Platform. That figure has climbed quickly. It was "more than 40" on October 13, and "more than 50" across Search and Gemini when Google opened the Connected Apps Platform preview on October 28, where developers integrate through a cloud-hosted MCP server or Android's AppFunctions APIs.
The school emails
The last example turns the inbox into an input channel. Woodward "forwarded a batch of school emails from Gmail to gemini@google.com. In less than a minute, every date was on my calendar with my wife invited." Gemini also "noticed a conflict during basketball practice and asked if I wanted to text one of the other parents to trade carpool duty."
Payments: AP2, tokens and a spending cap
The payment section is short, but it contains the post's most detailed commitments. "Gemini now uses the Agent Payments Protocol (AP2) to make purchases like this," it says. "We're working with American Express (coming soon), Mastercard, and Visa to keep your purchases secure with digital tokens on supported cards."
On rewards: "when you shop with Gemini, you still earn your card's points, cashback, or purchase protections during checkout." On control: "You can also set a hard spending cap upfront, and Gemini is designed to stay within your budget." The phrase "designed to" is a description of intent, not a guarantee, and the post does not say what happens if a purchase exceeds the cap.
AP2 is not new. When Google launched the Universal Commerce Protocol on January 11, 2026, co-developed with Shopify, Etsy, Wayfair, Target and Walmart, it presented AP2 as the companion protocol for cryptographic authorization of agent transactions. In March, Mastercard and Google introduced Verifiable Intent, a trust layer aligned with AP2 and UCP. On April 28, Google donated AP2 to the FIDO Alliance, and version 0.2 added support for payments made while the user is not present.
One sentence in today's post looks ahead: "Up next, we'll integrate the Universal Commerce Protocol (UCP) into Gemini so everyone can expect an even faster and more reliable shopping experience from discovery to checkout." This sits awkwardly beside PPC Land's October 15 report that Universal Cart, which runs on UCP, had already begun rolling out in the Gemini app in the US. The post does not explain how the two relate, or what UCP integration will add to Gemini.
The timing also places Google four days behind OpenAI. On November 13, OpenAI introduced ChatGPT Wallet, which lets its agents buy with a stand-in card credential. The two approaches differ in ways that matter to merchants and to users.
| Element | Gemini (Google, November 17) | ChatGPT Wallet (OpenAI, November 13) |
|---|---|---|
| Plan required | None named; task mode rolled out "free of charge" | ChatGPT Pro only |
| Geography | Task mode in six countries; payment geography not stated | US only, eligible US-issued cards |
| Card networks | Mastercard and Visa; American Express "coming soon" | American Express, Mastercard and Visa |
| Card protection | "digital tokens on supported cards" via AP2 | Stand-in card credential (network token) |
| Approval | Confirmation or pre-approval before "Buy"; "Always allow" for routine actions | Security check such as Face ID on every purchase |
| Spending limit | "hard spending cap upfront" | Delegated limits described as coming later |
Sources: today's Google post for Gemini; PPC Land's November 14 report for ChatGPT Wallet.
How a task runs, according to Google
The post describes the setup and the controls in enough detail to reconstruct the sequence.
- Start a task. According to Google, the user opens Gemini, flips "the home-screen toggle from Chat to Task," and describes what needs doing.
- Email route (US). Starting in the US, emails can be forwarded to gemini@google.com.
- Text route (US). The post gives four steps: open Gemini on Android or iOS and tap Settings; tap Channels; tap Set up under Text with Gemini; verify the phone number. A footnote adds: "You must have a US phone number with an eligible mobile carrier."
- During the task. "As Gemini works on a task, you can jump in to give it more instructions at any point, and it's also designed to pause and check with you before critical steps."
- Purchases. "Gemini is designed to ask for your confirmation or pre-approval before clicking 'Buy,' and will ask clarifying questions if an instruction is ambiguous." A hard spending cap can be set in advance.
- Routine actions. For actions the user already trusts, such as adding an item to a grocery cart, the post says the user "can set that to 'Always allow' so it doesn't pester you."
- Sites that resist. "On some sites, Gemini can make it most of the way through and then hand the screen back to you to click the final button."
- Afterwards. The user can "adjust task settings, disconnect Google and third-party apps, or delete chat history anytime," and failed tasks can be reported with a "thumbs down."
Where the work happens
Unlike Android task automation, which PPC Land reported runs inside an on-device virtual phone, Gemini tasks run in Google's cloud. "When you assign Gemini a task, it uses Google Cloud's Agent Compute Engine to work in a hardware-isolated virtual machine that's dedicated to your account and never shared with other users," according to the post. "Each virtual machine is sandboxed, so Gemini works in a separate, isolated environment that's regularly tested by our security red teams."
Google Cloud described the system in more depth on October 22. In that post, two Distinguished Engineers wrote that each agent session runs in a dedicated, single-tenant, memory-safe microVM built on Google's Titan and Titanium hardware, with no standard network interface inside the guest and no secrets stored there. At that point Google Cloud said Agent Compute Engine was targeted for preview in the coming months. Google's consumer app is therefore running on infrastructure that, as of October 22, Cloud customers could not yet use. Today's post does not say whether the preview has opened.
The cloud virtual machine approach mirrors a choice other companies have made. When Meta put its Muse agent inside WhatsApp on September 8, it also ran each agent on its own cloud virtual machine with its own browser.
The sandbox, Google says, is "just one of multiple layers of defense." The others named in the post are filters for prompt injection that scan search results, webpages and other content, and Google Safe Browsing, used "to limit which links Gemini can open." A companion technical post linked from the announcement, by Gautham Thambidorai, Vice President and Google Fellow, lists five layers: the microVMs, multi-stage prompt-injection defenses, a component called Policy Guardian that vets code before it runs, an Action Safety Agent that checks actions in real time and requires explicit confirmation before high-impact ones, and live browser handoffs where a human takes over sensitive steps. Google also invites researchers to test the system through its AI Vulnerability Reward Program.
There is a broader claim beneath the security language. "If your emails, calendar, passwords, photos, files, and payment cards are already in the Google ecosystem, Gemini is your best choice as a personal assistant because this sensitive data already lives at Google securely, and you don't have to send it elsewhere," Woodward writes. It is a pitch built on data concentration. It also arrives a month after Google changed its privacy policy, effective October 16, so that content from Gmail, Drive and Photos made available to other Google services through Personal Intelligence can be used to personalize ads, adding an exception to a pledge Google made in 2021. Today's post does not address advertising at all.
What Google says still goes wrong
The post is candid about failure. "While we've tested tasks extensively to minimize mistakes, we haven't eliminated all of them," it says. "You may encounter situations where a task fails to complete or misses the mark."
The main obstacle it names is the web itself. "For decades, bots were blocked by default because most were malicious. Personal assistants require a more nuanced approach, and we believe open, verifiable standards will benefit everyone." Google says it is "partnering directly with major websites and security providers," without naming any.
Hours before today's announcement, Google published a guide for site operators titled "Allow Google-Agent traffic." PPC Land reported earlier today that the guide lists six ways sites still block Google's agent after trying to allow it, from rule ordering and CAPTCHAs to rate limits on IP addresses that many users share. Google-Agent has been documented as a user-triggered fetcher since March 20. The guide and today's launch make the same point from two directions: Google wants sites to let its agent through, and it has now sent that agent to a much larger audience.
Site operators have raised the cost question. In a post published on November 5, Cloudflare argued that websites bear nearly all of 14 problems that AI agents create, because agents break the link between a site and the person behind each visit. Nothing in today's post addresses who pays for the extra traffic, or who is liable when an agent buys the wrong item.
Why this matters for marketers
Scale changes the calculation. Alphabet reported that the Gemini app passed 1 billion monthly active users in the third quarter, up from 650 million a year earlier. Alphabet does not disclose how many of those users pay. A feature limited to $19.99-and-up plans reached a fraction of that base; a free feature in six large markets, including the US and India, reaches far more of it.
For local businesses, the inbound call is the first contact point. Gemini picked "the nearest shops on Google Maps" and booked with the one that took appointments. Business Profile data, including location, hours and booking policies, therefore becomes the input an agent uses to decide who gets called. Salesforce argued in an October essay that personal agents will add contact-center traffic rather than reduce it, a vendor view that the essay did not support with data.
For retailers, the charger example describes agentic commerce at its most compressed: comparison through the Shopping Graph, clarifying questions, discount code and checkout in a single conversation. The merchant's product data feeds the comparison. The agent, not the shopper, decides which offers make the shortlist. Card rewards survive, according to Google, which removes one reason a shopper might prefer a retailer's own site.
Paid media is absent. The Gemini app carries no advertising, and Google's vice president of global ads, Dan Taylor, told Business Insider in January that there were no plans for ads in the Gemini app. Today's post mentions no sponsored results. That puts more purchase steps inside a surface where brands cannot buy placement, for now.
Competition is tightening around the same model. OpenAI's wallet is limited to its most expensive plan. Meta's Muse is free but US-only. Google's version is free in six countries and is wired into Gmail, Calendar, Maps, Password Manager and Wallet. The Open Markets Institute, an anti-monopoly group, argued in a 71-page report covered by PPC Land on November 9 that Google, Amazon, Microsoft and Meta are best placed to dominate AI agents because they already control the surrounding services. Woodward's own pitch, that Gemini is the "best choice" because the data "already lives at Google," describes that same advantage from the opposite side.
What the post does not say
Google gives no usage limits for free users, no age requirement, no full list of the "advanced features," and no country list for payments. It does not say whether paid subscribers keep Argon for tasks, when American Express support or UCP integration will arrive, how many businesses have opted out of calls, what merchants pay, if anything, or who carries liability when an agent makes a wrong purchase. The UK and the EU are not mentioned.
Key facts
- Source: "Put Gemini to work for you" (browser title "How to use Tasks in Gemini for personal assistance"), Google Keyword blog, Josh Woodward, Vice President, Labs, Gemini app, and AI Studio
- Published: November 17, 2026, 17:00 UTC
- URL: https://blog.google/innovation-and-ai/products/gemini-app/tasks-in-gemini-free-new-features/
- Model: Gemini 4 Hydrogen
- Free rollout countries: US, Japan, Korea, Canada, Australia, India
- Entry point: home-screen toggle from Chat to Task
- US channels: email forwarding to gemini@google.com; Text with Gemini via Settings, Channels, Set up, phone number verification
- Calling exclusions (footnote 1): Indiana, Louisiana, Minnesota, Montana, Nebraska
- Texting requirement (footnote 2): US phone number with an eligible mobile carrier
- Payments: Agent Payments Protocol (AP2); Mastercard and Visa; American Express "coming soon"; digital tokens on supported cards; hard spending cap set in advance
- Connected apps: more than 60; developers apply through the Connected Apps Platform
- Infrastructure: Google Cloud Agent Compute Engine, hardware-isolated virtual machine per account
- Security research: Google AI Vulnerability Reward Program (AI VRP)
- Companion announcement: "Google Wallet is now available on iOS in the U.S.," P.J. Linarducci, November 17, 2026
- Business opt-out: Google Business Profile Help, "About automated calls and texts from Google to your business"
- Next step named: Universal Commerce Protocol integration into Gemini, no date
Timeline
- July 16, 2025 - Google adds AI-powered business calling to US Search, excluding Indiana, Louisiana, Minnesota, Montana and Nebraska
- November 13, 2025 - Google launches "Let Google Call" and agentic checkout with Google Pay
- January 11, 2026 - Google launches the Universal Commerce Protocol alongside AP2
- January 2026 - Dan Taylor says there are no plans for ads in the Gemini app
- March 2026 - Mastercard and Google introduce Verifiable Intent, aligned with AP2 and UCP
- March 20, 2026 - Google documents Google-Agent as a user-triggered fetcher
- April 28, 2026 - Google donates AP2 to the FIDO Alliance; version 0.2 adds payments without the user present
- July 30, 2026 - Gemini Spark expands to 160-plus countries but stays unavailable in the EEA, UK, Switzerland and Nigeria
- September 8, 2026 - Meta puts its Muse agent, running on cloud virtual machines, inside WhatsApp in the US
- October 4, 2026 - Google's plan documentation limits free users to Flash-Lite from October 9 and lists AI Pro at $19.99
- October 13, 2026 - Gemini app adds a chat-to-tasks toggle in a post naming AI Pro and Ultra subscribers
- October 15, 2026 - Universal Cart begins rolling out in US Search and the Gemini app
- October 16, 2026 - Google privacy policy allows Personal Intelligence data to shape ads
- October 22, 2026 - Google Cloud describes Agent Compute Engine, targeted for preview in the coming months
- October 28, 2026 - Google opens the Connected Apps Platform preview with more than 50 apps across Search and Gemini
- October 28, 2026 - Alphabet reports more than 1 billion monthly Gemini app users for the third quarter
- November 5, 2026 - Cloudflare argues websites bear nearly all of 14 AI agent problems
- November 9, 2026 - Open Markets Institute report on Big Tech and AI agents covered by PPC Land
- November 10, 2026 - Android task automation reaches the UK, India and Japan
- November 12, 2026 - Gemini 4 Hydrogen reaches Google Search for all users
- November 13, 2026 - OpenAI introduces ChatGPT Wallet for Pro subscribers in the US
- November 16, 2026 - Ask Maps starts calling US salons, groomers and garages
- November 16, 2026 - Google's holiday guide steers bookings into AI Mode; monitoring tasks name no paid tier
- November 17, 2026 - Google publishes "Allow Google-Agent traffic" for site operators
- November 17, 2026 - Google Wallet becomes available on iOS in the US
- November 17, 2026 - Gemini task mode rolls out free in six countries, with calls, texting, email forwarding and AP2 payments starting in the US
Related PPC Land coverage
- Gemini app tasks mode works while users are away, for $19.99 and up - the October 13 launch of the chat-to-tasks toggle for AI Pro and Ultra subscribers.
- Google Search gains Gemini 4 Hydrogen for all users, not just subscribers - the model behind today's free rollout, with pricing and performance claims.
- Google cuts two of three Gemini models for free users from October 9 - Gemini plan prices and the free-tier model restriction.
- Google's Gemini can now use food and ride apps for people in 3 more countries - Android task automation and its on-device virtual phone.
- Google Maps' AI now calls US salons, groomers and garages to find open slots - Ask Maps adds phone calls to local businesses.
- Google search calls small businesses for customers automatically - the 2025 launch of AI calling in Search and its five-state exclusion.
- Google launches agentic checkout and AI shopping tools for holiday season - "Let Google Call" and Google Pay checkout in 2025.
- Google launches protocol for AI agents to shop across platforms - the UCP launch and the role of AP2.
- FIDO Alliance forms working groups to lock down AI agent payments - Google's donation of AP2 and the version 0.2 changes.
- Mastercard and Google's new trust layer could reshape how AI buys for you - Verifiable Intent and its link to AP2.
- OpenAI hands its AI agents a wallet without the real card number in it - ChatGPT Wallet's terms, four days before Google's announcement.
- Google's Universal Cart reaches US Search and Gemini, but not YouTube - the cart that already runs on UCP inside Gemini.
- Google opens preview for third-party apps inside Search and Gemini - the Connected Apps Platform and its app counts.
- Google lists six ways sites still block its AI agent after allowing it - Google's guide for letting Google-Agent through CDNs and firewalls.
- Google-Agent joins the crawler list as AI browsing gets an official identity - the agent's user-agent identity and IP ranges.
- Websites pay for nearly all 14 AI agent problems, Cloudflare argues - the cost of agent traffic for site operators.
- Google lets connected Gmail and Photos data shape ads, 5 years after pledge - the October privacy policy change tied to Personal Intelligence.
- Gemini app gains 350m users in a year to pass 1bn, Alphabet says - Gemini app scale in the third quarter.
- Gemini Spark blocks EU and UK users as Google adds 160 countries - Google's earlier agent and its European exclusions.
- Meta puts an AI agent that buys things inside WhatsApp, US only - Meta's Muse and its cloud virtual machine design.
- Google, Amazon, Microsoft, Meta could dominate AI agents, Open Markets warns - an anti-monopoly group's case on agent gatekeepers.
- Why Google won't put ads in Gemini - VP explains the strategy gap - Google's stated position on ads in the Gemini app.
- Salesforce says personal AI agents will add call center traffic, not cut it - a vendor view on what consumer agents mean for business phone lines.
- Google steers holiday hotel, dinner and ticket bookings into AI Mode - AI Mode tasks and bookings a day before today's post.
Summary
Who: Google, through a blog post by Josh Woodward, Vice President of Labs, the Gemini app and AI Studio. Partners named include Mastercard, Visa, American Express (coming soon) and Instacart. Those affected include Gemini app users, local businesses that receive calls, retailers whose products Gemini compares and buys, and website operators whose sites the agent visits.
What: The Gemini app's task mode, introduced on October 13 in a post naming AI Pro and Ultra subscribers, is rolling out free of charge on Gemini 4 Hydrogen. US users also get calls to businesses, texting over RCS, email forwarding to gemini@google.com and purchases through the Agent Payments Protocol with tokenized cards and a spending cap. Tasks run in isolated cloud virtual machines on Google Cloud's Agent Compute Engine.
When: Today, November 17, 2026, with the post published at 17:00 UTC. American Express support and UCP integration have no dates.
Where: The free task mode covers the US, Japan, South Korea, Canada, Australia and India. Calling, texting and email forwarding start in the US, and calling is unavailable in Indiana, Louisiana, Minnesota, Montana and Nebraska. No European country is listed.
Why: Google says its goal is for Gemini "to save you time and money," and argues that users whose data already sits in Google's services should keep it there. For marketers, the change moves calls, comparisons, discount codes and checkouts into an ad-free agent that reaches more than 1 billion monthly users, while Google concedes the agent still makes mistakes and that many sites are not yet built to let it through.
Discussion