Google Assistant Shutdown: Gemini Takes Over Android

The most widely deployed voice assistant in the Android world is being switched off. On September 4, 2026, Google began removing Google Assistant from Android devices and replacing it with Gemini, the company’s AI model, in a transition that had been signaled for weeks and is now underway. Google notified users by email that Assistant is “being discontinued,” describing Gemini as “the assistant experience on Android” going forward.

The change is not a flip of a single switch. Google says the removal will roll out over several weeks, so devices will lose Assistant on different days rather than all at once. But the direction is fixed: once the transition reaches a given device, its owner cannot switch back to the old Assistant.

What is changing, and where

Starting September 4, Google is removing Assistant and installing Gemini across a broad set of Android hardware: phones, tablets, Wear OS watches, headphones, and Android Auto. These are the surfaces where hundreds of millions of people have summoned Assistant with a wake word or a long-press for the better part of a decade.

Not every product is included in this wave. Google says the change does not yet cover cars with Google built-in, Google TV, or Home and Nest speakers. Those platforms keep Assistant for now, which leaves users in the odd interim position of talking to Gemini on their phone and Assistant on their living-room speaker. Google has indicated the assistant’s various surfaces will migrate over time, but the September rollout is focused on the mobile and wearable tier.

For most people the visible change is what answers when they invoke the assistant. The wake phrase and the press-and-hold gesture stay familiar; the intelligence behind them is replaced. Where Assistant was a command-and-response system tuned over years to handle a fixed catalog of tasks, Gemini is a general-purpose AI model — more capable at open-ended conversation and reasoning, and, Google argues, better suited to the kind of multi-step requests that Assistant struggled with.

Gemini Spark and the agent pitch

The headline capability Google is using to justify the swap is agentic behavior. Alongside the transition, Google has been promoting Gemini Spark, an AI agent that the company says can manage tasks more efficiently than Assistant by acting on a user’s behalf rather than simply answering.

With a user’s permission, Google says Gemini Spark can access logged-in accounts and saved passwords to complete multi-step tasks across the web — the examples cited include scheduling appointments, filling out forms, and carrying out routine online activities. That is a meaningful step beyond what Assistant did. Assistant could set a timer, answer a trivia question, or control a smart bulb; an agent that can log in and complete a booking is operating in a different category, closer to a delegate than a voice remote.

It is also a category that raises the stakes on trust. An assistant that can reach into accounts and use stored credentials to act autonomously is powerful precisely because it holds the keys, and the permission model around that access — what it can touch, when it asks, and how a user revokes it — becomes the whole ballgame. Google frames the access as permissioned and user-granted, but the shift from answering questions to taking actions is the part users will feel, for better and worse.

What gets lost in the switch

A replacement is rarely a perfect superset of what it replaces, and Google has been candid that some Assistant capabilities may not have direct Gemini equivalents at launch. The company is advising users to check a handful of features before their device makes the jump, specifically calling out routines, reminders, Interpreter Mode, and media controls.

Those are not fringe features. Routines automate strings of actions; reminders are among the most-used assistant functions; Interpreter Mode provided real-time translation; and media controls are how many people drive playback hands-free. Users who rely on any of them are being told, in effect, to verify that their setup survives the migration — and to do so before the change lands, since there is no going back afterward.

Beyond the feature gaps, there is a performance question. Some users testing Gemini in place of Assistant have reported that it can be less reliable on basic, mechanical tasks — the quick, deterministic “set a 10-minute timer” or “turn off the lights” requests that Assistant executed instantly. A large language model is a different tool than a command parser: it is far more flexible and far more conversational, but that generality can introduce latency or unpredictability on the simple chores where the old system was fast and dependable. Whether Gemini closes that gap as it matures on-device is one of the open questions of the transition.

A milestone in Google’s all-in AI shift

The Assistant shutdown is the clearest signal yet of how completely Google has reorganized around Gemini. The brand has moved from a chatbot to the connective tissue of Google’s consumer products, reaching the one-billion-user milestone earlier this year and threading through the company’s search experience via AI Mode. Retiring Assistant folds the last major standalone consumer surface into that single model family, built on the Gemini 3 generation that anchors Google’s current lineup.

The move also lands against a regulatory backdrop that is watching exactly this kind of consolidation. Google’s control over the default assistant on Android has drawn scrutiny in Europe, where competition authorities have pressed the company over how it surfaces AI rivals on the platform, as covered in our report on the EU’s DMA order over Google, Android, and AI rivals. Swapping the entire base of Android devices onto Google’s own model — one that can act on users’ accounts — is the sort of default-setting power regulators have signaled they intend to police.

What it means

For the roughly billion-plus people who use Android, this is the moment the AI transition stops being an app they can open and becomes the assistant baked into their phone. Most will experience it as an upgrade in conversational ability and a downgrade in muscle memory: Gemini can hold a real dialogue and, through Spark, actually do things, but the reflexive, instant reliability of the old Assistant is exactly what some early users say they miss.

The practical advice is simple and time-sensitive. Because the change is irreversible per device and rolls out on an unpredictable schedule, anyone who depends on Assistant routines, reminders, translation, or hands-free media control should audit those workflows now and rebuild them in Gemini where the equivalents exist. Waiting until Gemini arrives means discovering the gaps after the door has closed.

The bigger story is strategic. By replacing Assistant wholesale, Google is betting that an agentic AI that can log in and complete tasks is worth the disruption of retiring a mature, trusted product — and it is placing that bet on the most personal computing surface people own. If Gemini Spark delivers on the promise of a genuine digital delegate, the friction of the switch will be a footnote. If it stumbles on the basics while reaching for the ambitious, Google will have spent a decade of accumulated Assistant goodwill to find out. Either way, the era of Google Assistant is ending, and the era of Gemini as the assistant has begun.