Google Assistant stops working on Android phones, tablets, watches and headphones starting September 4, and Gemini takes its place. If you have ever said “Hey Google” to set a timer or turn off a lamp, the thing answering you is about to change, and you will not be able to switch back.
The Gist
- Google Assistant shuts down from September 4 on phones, tablets, Wear OS watches, headphones and Android Auto.
- Gemini replaces it, and once your device switches over you cannot go back to the old assistant.
- Two features are being dropped: live conversation translation and the daily smart home briefing.
- Gemini adds roughly half a second of thinking time before it acts on a spoken command.
Have ChatGPT Recap This Article
ChatGPTWhat actually happens on September 4
Google has set a date for something it has been hinting at for two years. Starting September 4, Google Assistant stops working on Android phones and tablets, on Wear OS watches, on headphones that support it, and in cars running Android Auto. Google laid out the plan in its own note explaining how the mobile assistant is being upgraded to Gemini.
You will not all get it on the same day. The rollout starts on September 4 and takes a few weeks to reach everyone, device by device. There is no button to speed it up and no button to delay it.
One detail matters more than the rest. Once your device switches, the change is permanent. You cannot restore Google Assistant on that phone, tablet or paired accessory afterwards. The only exception is cars with Google built directly into the dashboard, which keep Assistant for now.
How Gemini answers a request differently
So what is actually replacing Assistant? Gemini is Google’s AI chat, the same kind of tool as ChatGPT. It writes its answer one word at a time, thinking as it goes, rather than matching your sentence to a fixed list of commands.
That gives you something Assistant never handled well. You can ask a messy question, follow up without repeating yourself, and ask it to reason through something instead of just fetching a fact. It is the same engine that Google has been putting everywhere, including the morning briefing that reads your inbox and sums it up.
There is a cost to that flexibility, and it shows up as waiting. Gemini adds between 380 and 620 milliseconds of thinking time before it even starts doing what you asked. On a real question you will never notice. On “turn off the kitchen light” shouted from the next room, half a second of silence feels long.
Two features do not survive the move at all. Interpreter Mode, which translated a live conversation between two people, is gone. So is the daily smart home briefing that told you what was happening around the house.
Keep learning on AI Noobies:
- Reward Hacking: Why AI Agents Cheat to Win
- New EU AI Rules Make Chatbots Say They’re AI
- MiniMax H3 Is a New Chinese AI Video Model
What “deterministic” and “probabilistic” really mean here
Those two words explain the whole story, and they are simpler than they sound. A deterministic system always gives the same output for the same input. Say the exact phrase, get the exact action, every single time.
Google Assistant worked that way. It recognised a phrase from a list and fired off the matching command. Boring, limited, and extremely reliable for small jobs like timers, alarms and lights.
A probabilistic system builds its answer by predicting what comes next, which means it can handle phrasing it has never seen, and it can also surprise you. Ask the same thing twice and you may get two slightly different responses. That is normal for an AI chat, and it is new territory for a light switch.
Zooming out, this is the argument you will start hearing in the wider conversation about AI. Everyday tools built on old, predictable software are being rebuilt on models that are smarter and less certain. Apple is running the same swap on its side, which we unpacked in our explainer on the new Siri that finally holds a conversation. Voice assistants are simply the first place where millions of people will feel that trade in their own kitchen.
How to get your voice habits ready this month
You have a few weeks, so there is a useful thing to do with them. Make a short list of the voice commands you actually use, the three or four you say without thinking, and note exactly how you phrase them.
If you rely on Interpreter Mode when you travel, start looking for a translation app now rather than discovering the gap at an airport. If your morning depends on the smart home briefing, that habit needs a replacement too.
You can also try Gemini before it becomes your only option. It is already available on Android, and the cheaper, faster version described in our piece on Gemini 3.6 Flash is what most people will be talking to. Spend a week asking it the things you normally ask Assistant and you will know what to expect.
One last thing worth knowing: nothing you have to do here is urgent, but nothing you skip gets easier later either. The switch arrives on its own schedule, and the people who spend twenty minutes testing their own commands in August will be the ones who barely notice September.
Stay tuned on AI Noobies.
About the Author
