Google opened Gemini in Chrome to all Android customers in the USA on August 18, 2026, bringing its built-in searching assistant to the complete U.S. Android person base for the primary time. Introduced on The Key phrase by Charmaine Dsilva, Director of Product Administration for Chrome, the discharge additionally extends auto browse, the browser’s agentic task-completion function, to telephones for Google AI Professional and AI Extremely subscribers within the U.S.
Gemini in Chrome summarizes lengthy articles, solutions questions in regards to the web page a person is studying, and connects to Google apps together with Calendar and Preserve with out requiring a change between tabs. The Android model additionally contains picture technology and modifying on the undergo Nano Banana. Customers attain the assistant by the Gemini icon or the three-dot menu on Chrome’s prime bar on an Android machine.
The auto browse enlargement is the extra consequential piece for subscribers. Slightly than answering questions on a web page, auto browse carries out multi-step duties on the net on the person’s behalf: reserving parking for an occasion, updating a recurring on-line order, or organizing journey for a trip. Till now, the function ran on desktop Chrome solely; the August 18 announcement places it on Android for paying tiers.
“These options are constructed to be safe by design, and our fashions are educated to detect recognized threats, equivalent to immediate injection,” Dsilva wrote. “And for added management, auto browse is designed to ask for affirmation earlier than finishing some delicate duties.”
How Auto Browse Works As soon as It Is Operating
The mechanism behind the function is documented in Google’s Chrome assist pages. When a person describes a job, Gemini in Chrome proposes a plan first; the person critiques that plan, confirms the web page context and any private info concerned, and begins the duty explicitly. Gemini then chooses which internet sites to make use of, and a visual auto browse icon marks the tab it’s working in. The person can take over at any level, hand the duty again, or cease it outright.
The plan-review step exists as a result of the agent operates with broad attain. Auto browse can act throughout the websites a person is signed into, and it could draw on private info from linked Google apps to finish duties, sharing that info with the websites it visits. With separate permission, it might probably signal into websites utilizing saved credentials by Google Password Supervisor, which Google says doesn’t share the passwords themselves with Gemini.
For delicate steps, the system is designed to pause slightly than proceed. Google’s documentation says the agent asks for evaluate and affirmation earlier than actions together with sending communications, submitting internet kinds, scheduling occasions, and accessing websites carrying monetary or well being knowledge, and it could hand management again to the person for steps like finalizing transactions or accepting phrases of service. It additionally refuses a class of prohibited duties and restricts its exercise to websites and actions related to the request.
What the Tier Gate Buys, and The place the Limits Sit
Auto browse requires a Google AI Professional or AI Extremely subscription on a private account, customers have to be 18 or over and within the U.S. with their machine language set to English, and the function doesn’t run in Incognito mode. Every day caps scale with the plan: AI Professional subscribers rise up to twenty multi-step job requests per day, whereas AI Extremely subscribers rise up to 200, with a restrict on what number of duties can run on the similar time. Google is releasing the function progressively, so even eligible subscribers might not see it instantly.
The safety framing behind these limits was specified by a December 8, 2025 put up from the Chrome safety group, which recognized oblique immediate injection, the place malicious directions hidden in internet content material steer an agent towards unintended actions, as the first new menace dealing with agentic browsers. Google’s layered response features a separate “person alignment critic” mannequin that vets every proposed motion earlier than execution, origin restrictions that confine the agent to task-relevant websites, a real-time prompt-injection classifier, and a bug bounty of as much as $20,000 for demonstrated breaches of the agentic safety boundaries.
Google’s personal documentation is direct in regards to the residual danger: auto browse is an experimental function, Gemini in Chrome could make errors or act unexpectedly, together with finishing a purchase order with out permission, and the safeguards don’t assure safety in opposition to all dangers. Customers stay answerable for the agent’s actions throughout a job, which is why the plan evaluate and affirmation prompts are the load-bearing controls.
The Street to a Full Android Launch
The cellular basic availability introduced August 18 is the endpoint of a staged rollout. Google first introduced Gemini in Chrome to desktop customers within the U.S. in September 2025, on the time promising that agentic capabilities would comply with, and previewed auto browse earlier than delivery it on desktop for paid tiers. The safety structure arrived alongside that desktop preview, and Unite.AI coated the announcement when Google detailed the agentic safeguards for Chrome’s AI options in December 2025.
The assistant itself is now one among Google’s widest-reach AI surfaces. The standalone Gemini app crossed 1 billion month-to-month customers earlier in August 2026, and embedding it within the default browser on Android extends that attain to the context the place cellular customers already spend their searching time.
What ships subsequent is basically a matter of widening the identical gates: Google’s assist pages observe that auto browse will not be out there in Reside chats or in Gemini in Chrome on iPhone and iPad, and that work and faculty accounts want administrator enablement. The Android launch covers U.S. customers in English; the desktop model has already expanded to further areas, and the gradual cellular launch is continuous to roll out to eligible accounts.
